diff --git a/.drone/drone.yml b/.drone/drone.yml deleted file mode 100644 index 8476f7f..0000000 --- a/.drone/drone.yml +++ /dev/null @@ -1,165 +0,0 @@ -kind: pipeline -type: exec -name: linux-amd64 - -platform: - arch: amd64 - os: linux - -steps: -- name: test - commands: - - mkdir -p $HOMEPATH/.docker; echo $IMAGE_PULL_SECRETS > $HOMEPATH/.docker/config.json - - cd grafana && make test && cd .. - - cd telegraf && make test && cd .. - environment: - VERSION: ${DRONE_TAG:-latest}-linux-amd64 - DEV_REGISTRY: - from_secret: dev_registry - DRYCC_REGISTRY: - from_secret: drycc_registry - IMAGE_PULL_SECRETS: - from_secret: container_pull_secrets - when: - event: - - push - - tag - - pull_request - -- name: publish - commands: - - echo $CONTAINER_PASSWORD | docker login $DRYCC_REGISTRY --username $CONTAINER_USERNAME --password-stdin > /dev/null 2>&1 - - cd grafana && make build-all push-all - environment: - VERSION: ${DRONE_TAG:-latest}-linux-amd64 - DEV_REGISTRY: - from_secret: dev_registry - DRYCC_REGISTRY: - from_secret: drycc_registry - CONTAINER_USERNAME: - from_secret: container_username - CONTAINER_PASSWORD: - from_secret: container_password - when: - event: - - push - - tag - ---- -kind: pipeline -type: exec -name: linux-arm64 - -platform: - arch: arm64 - os: linux - -steps: -- name: publish - commands: - - echo $CONTAINER_PASSWORD | docker login $DRYCC_REGISTRY --username $CONTAINER_USERNAME --password-stdin > /dev/null 2>&1 - - cd grafana && make build-all push-all - environment: - VERSION: ${DRONE_TAG:-latest}-linux-arm64 - DEV_REGISTRY: - from_secret: dev_registry - DRYCC_REGISTRY: - from_secret: drycc_registry - CONTAINER_USERNAME: - from_secret: container_username - CONTAINER_PASSWORD: - from_secret: container_password - when: - event: - - push - - tag - ---- -kind: pipeline -type: docker -name: manifest -image_pull_secrets: -- container_pull_secrets - -steps: -- name: generate grafana manifest - image: registry.drycc.cc/drycc/python-dev - pull: always - commands: - - sed -i "s/registry.drycc.cc/$${DRYCC_REGISTRY}/g" .drone/manifest.tmpl - - sed -i "s/monitor/grafana/g" .drone/manifest.tmpl - environment: - DRYCC_REGISTRY: - from_secret: drycc_registry - -- name: publish grafana - image: plugins/manifest - settings: - spec: .drone/manifest.tmpl - username: - from_secret: container_username - password: - from_secret: container_password - environment: - DEV_REGISTRY: - from_secret: dev_registry - DRYCC_REGISTRY: - from_secret: drycc_registry - -- name: generate telegraf manifest - image: registry.drycc.cc/drycc/python-dev - pull: always - commands: - - sed -i "s/registry.drycc.cc/$${DRYCC_REGISTRY}/g" .drone/manifest.tmpl - - sed -i "s/grafana/telegraf/g" .drone/manifest.tmpl - environment: - DRYCC_REGISTRY: - from_secret: drycc_registry - -- name: publish telegraf - image: plugins/manifest - settings: - spec: .drone/manifest.tmpl - username: - from_secret: container_username - password: - from_secret: container_password - environment: - DEV_REGISTRY: - from_secret: dev_registry - DRYCC_REGISTRY: - from_secret: drycc_registry - -trigger: - event: - - push - - tag - -depends_on: -- linux-amd64 -- linux-arm64 - ---- -kind: pipeline -type: exec -name: chart - -steps: -- name: generate chart - commands: - - IMAGE_TAG=$([ ! -z $DRONE_TAG ] && echo \"${DRONE_TAG:1}\" || echo \"canary\") - - sed -i "s/imageTag:\ \"canary\"/imageTag:\ $IMAGE_TAG/g" charts/monitor/values.yaml - - helm package -u charts/monitor --version $([ -z $DRONE_TAG ] && echo 1.0.0 || echo ${DRONE_TAG#v}) - - echo $CONTAINER_PASSWORD | helm registry login $DRYCC_REGISTRY -u $CONTAINER_USERNAME --password-stdin - - helm push monitor-$([ -z $DRONE_TAG ] && echo 1.0.0 || echo ${DRONE_TAG#v}).tgz oci://$DRYCC_REGISTRY/$([ -z $DRONE_TAG ] && echo charts-testing || echo charts) - environment: - DRYCC_REGISTRY: - from_secret: drycc_registry - CONTAINER_USERNAME: - from_secret: container_username - CONTAINER_PASSWORD: - from_secret: container_password - when: - event: - - push - - tag diff --git a/.drone/manifest.tmpl b/.drone/manifest.tmpl deleted file mode 100644 index 6080b08..0000000 --- a/.drone/manifest.tmpl +++ /dev/null @@ -1,18 +0,0 @@ -image: registry.drycc.cc/drycc/monitor:{{#if build.tag}}{{trimPrefix "v" build.tag}}{{else}}canary{{/if}} -{{#if build.tags}} -tags: -{{#each build.tags}} - - {{this}} -{{/each}} -{{/if}} -manifests: - - - image: registry.drycc.cc/drycc/monitor:{{#if build.tag}}{{build.tag}}-{{else}}latest-{{/if}}linux-amd64 - platform: - architecture: amd64 - os: linux - - - image: registry.drycc.cc/drycc/monitor:{{#if build.tag}}{{build.tag}}-{{else}}latest-{{/if}}linux-arm64 - platform: - architecture: arm64 - os: linux diff --git a/.gitignore b/.gitignore index 72d9643..81b005a 100644 --- a/.gitignore +++ b/.gitignore @@ -2,8 +2,5 @@ *.swp *.swo .DS_Store - -telegraf/manifests/*.tmp.yaml -chronograf/manifests/*.tmp.yaml -grafana/manifests/*.tmp.yaml -rootfs/usr/bin/telegraf +.sisyphus +__pycache__ \ No newline at end of file diff --git a/.woodpecker/build-linux.yml b/.woodpecker/build-linux.yml new file mode 100644 index 0000000..585a87b --- /dev/null +++ b/.woodpecker/build-linux.yml @@ -0,0 +1,35 @@ +matrix: + platform: + - linux/amd64 + - linux/arm64 + +labels: + type: exec + platform: ${platform} + +steps: +- name: publish-linux + image: bash + commands: + - echo $CONTAINER_PASSWORD | podman login $DRYCC_REGISTRY --username $CONTAINER_USERNAME --password-stdin > /dev/null 2>&1 + - export VERSION=$([ -z $CI_COMMIT_TAG ] && echo latest || echo $CI_COMMIT_TAG)-$(sed 's#/#-#g' <<< $CI_SYSTEM_PLATFORM) + - make podman-build podman-push + environment: + CODENAME: + from_secret: codename + DEV_REGISTRY: + from_secret: dev_registry + DRYCC_REGISTRY: + from_secret: drycc_registry + CONTAINER_USERNAME: + from_secret: container_username + CONTAINER_PASSWORD: + from_secret: container_password + when: + event: + - push + - tag + - cron + +depends_on: +- test-linux diff --git a/.woodpecker/chart.yaml b/.woodpecker/chart.yaml new file mode 100644 index 0000000..cfcd3ec --- /dev/null +++ b/.woodpecker/chart.yaml @@ -0,0 +1,33 @@ +labels: + type: exec + platform: linux/amd64 + +steps: +- name: generate-chart + image: bash + commands: + - export VERSION=$(sed 's#v##' <<< $CI_COMMIT_TAG) + - export IMAGE_TAG=$([ ! -z $CI_COMMIT_TAG ] && echo \"$VERSION\" || echo \"canary\") + - export APP_VERSION=$([ -z $CI_COMMIT_TAG ] && echo $CI_COMMIT_SHA || echo $VERSION) + - export CHART_VERSION=$([ -z $CI_COMMIT_TAG ] && echo 1.0.0 || echo $VERSION) + - sed -i "s/imageTag:\ \"canary\"/imageTag:\ $IMAGE_TAG/g" charts/$${CI_REPO_NAME}/values.yaml + - helm package -u charts/$${CI_REPO_NAME} --version $CHART_VERSION --app-version $APP_VERSION + - echo $CONTAINER_PASSWORD | helm registry login $DRYCC_REGISTRY -u $CONTAINER_USERNAME --password-stdin + - helm push $${CI_REPO_NAME}-$CHART_VERSION.tgz oci://$DRYCC_REGISTRY/$([ -z $CI_COMMIT_TAG ] && echo charts-testing || echo charts) + environment: + DEV_REGISTRY: + from_secret: dev_registry + DRYCC_REGISTRY: + from_secret: drycc_registry + CONTAINER_USERNAME: + from_secret: container_username + CONTAINER_PASSWORD: + from_secret: container_password + when: + event: + - push + - tag + - cron + +depends_on: +- manifest diff --git a/.woodpecker/manifest.tmpl b/.woodpecker/manifest.tmpl new file mode 100644 index 0000000..8739535 --- /dev/null +++ b/.woodpecker/manifest.tmpl @@ -0,0 +1,18 @@ +image: registry.drycc.cc/drycc/{{project}}:{{#if build.tag}}{{trimPrefix "v" build.tag}}{{else}}canary{{/if}} +{{#if build.tags}} +tags: +{{#each build.tags}} + - {{this}} +{{/each}} +{{/if}} +manifests: + - + image: registry.drycc.cc/drycc/{{project}}:{{#if build.tag}}{{build.tag}}-{{else}}latest-{{/if}}linux-amd64 + platform: + architecture: amd64 + os: linux + - + image: registry.drycc.cc/drycc/{{project}}:{{#if build.tag}}{{build.tag}}-{{else}}latest-{{/if}}linux-arm64 + platform: + architecture: arm64 + os: linux diff --git a/.woodpecker/manifest.yml b/.woodpecker/manifest.yml new file mode 100644 index 0000000..88ded66 --- /dev/null +++ b/.woodpecker/manifest.yml @@ -0,0 +1,43 @@ +labels: + type: exec + platform: linux/amd64 + +steps: +- name: generate-manifest + image: bash + commands: + - sed -i "s/{{project}}/$${CI_REPO_NAME}/g" .woodpecker/manifest.tmpl + - sed -i "s/registry.drycc.cc/$${DRYCC_REGISTRY}/g" .woodpecker/manifest.tmpl + environment: + DRYCC_REGISTRY: + from_secret: drycc_registry + when: + event: + - tag + - push + - cron + +- name: publish-manifest + image: bash + commands: + - podman run --rm + -e PLUGIN_SPEC=.woodpecker/manifest.tmpl + -e PLUGIN_USERNAME=$CONTAINER_USERNAME + -e PLUGIN_PASSWORD=$CONTAINER_PASSWORD + -e DRONE_TAG=$CI_COMMIT_TAG + -v $(pwd):$(pwd) + -w $(pwd) + plugins/manifest + environment: + CONTAINER_USERNAME: + from_secret: container_username + CONTAINER_PASSWORD: + from_secret: container_password + when: + event: + - tag + - push + - cron + +depends_on: +- build-linux diff --git a/.woodpecker/test-linux.yml b/.woodpecker/test-linux.yml new file mode 100644 index 0000000..a1f9f7d --- /dev/null +++ b/.woodpecker/test-linux.yml @@ -0,0 +1,25 @@ +matrix: + platform: + - linux/amd64 + - linux/arm64 + +labels: + type: exec + platform: ${platform} + +steps: +- name: test-linux + image: bash + commands: + - make test + environment: + CODENAME: + from_secret: codename + DEV_REGISTRY: + from_secret: dev_registry + when: + event: + - push + - tag + - cron + - pull_request diff --git a/Makefile b/Makefile new file mode 100644 index 0000000..b964ce2 --- /dev/null +++ b/Makefile @@ -0,0 +1,37 @@ +SHELL := /bin/bash + +# grafana share/use the following targets/exports +SHORT_NAME ?= grafana +SHELL_SCRIPTS = rootfs/usr/share/grafana/oauth2 + +BUILD_TAG ?= git-$(shell git rev-parse --short HEAD) +DRYCC_REGISTRY ?= ${DEV_REGISTRY} +IMAGE_PREFIX ?= drycc +PLATFORM ?= linux/amd64,linux/arm64 + +include includes.mk +include versioning.mk + +TEST_ENV_PREFIX := podman run --rm -v ${CURDIR}:/bash -w /bash ${DEV_REGISTRY}/drycc/python-dev + +build: podman-build +push: podman-push +deploy: check-kubectl podman-build podman-push install + +podman-build: + podman build --build-arg CODENAME=${CODENAME} -t ${IMAGE} rootfs + podman tag ${IMAGE} ${MUTABLE_IMAGE} + +podman-buildx: + podman buildx build --build-arg CODENAME=${CODENAME} --platform ${PLATFORM} -t ${IMAGE} rootfs --push + +clean: check-podman + podman rmi $(IMAGE) + +test: test-style + +test-style: + ${TEST_ENV_PREFIX} flake8 --show-source --max-line-length=100 $(SHELL_SCRIPTS) + +.PHONY: build push podman-build clean upgrade deploy test test-style + diff --git a/README.md b/README.md index 8c63d10..0376504 100644 --- a/README.md +++ b/README.md @@ -1,56 +1,41 @@ -# Drycc Monitor v2 +# Drycc Grafana -[![Build Status](https://drone.drycc.cc/api/badges/drycc/monitor/status.svg)](https://drone.drycc.cc/drycc/monitor) +[![Build Status](https://woodpecker.drycc.cc/api/badges/drycc/grafana/status.svg)](https://woodpecker.drycc.cc/drycc/grafana) -Drycc (pronounced DAY-iss) Workflow is an open source Platform as a Service (PaaS) that adds a developer-friendly layer to any [Kubernetes](http://kubernetes.io) cluster, making it easy to deploy and manage applications on your own servers. +Drycc (pronounced DAY-iss) Workflow is an open source Platform as a Service (PaaS) that adds a developer-friendly layer to any [Kubernetes][k8s-home] cluster, making it easy to deploy and manage applications on your own servers. For more information about the Drycc Workflow, please visit the main project page at https://github.com/drycc/workflow. We welcome your input! If you have feedback, please [submit an issue][issues]. If you'd like to participate in development, please read the "Development" section below and [submit a pull request][prs]. -# About -This repository aims to contain the necessary components for a production quality monitoring solution that runs on top of the kubernetes cluster scheduler. It provides part of the [TICK](https://influxdata.com/time-series-platform/) stack which is produced by the influxdata team. - -## Current State -Currently this repo provides only 2 components (Telegraf, and Grafana). Telegraf is the metrics collection agent that runs using the daemon set API. For more infomation please read [this](telegraf/README.md). - -Grafana is a stand alone graphing application. It natively supports Influxdb as a datasource and provides a robust engine for creating dashboards on top of timeseries data. We provide a few out of the box dashboards for monitoring Drycc Workflow and Kubernetes but please feel free to use them as a starting point for creating your own dashboards. - -# Architecture Diagram - -``` - ┌────────┐ - │ Router │ ┌────────┐ - └────────┘ │ Logger │ - │ └────────┘ - Log file │ - │ │ - ▼ ▼ -┌────────┐ ┌─────────┐ logs/metrics ┌──────────────┐ -│App Logs│──Log File──▶│ fluentd │───────topics─────▶│ Redis Stream │ -└────────┘ └─────────┘ └──────────────┘ - │ - │ -┌─────────────┐ │ -│ HOST │ ▼ -│ Telegraf │───┐ ┌────────┐ -└─────────────┘ │ │Telegraf│ - │ └────────┘ -┌─────────────┐ │ │ -│ HOST │ │ ┌───────────┐ │ -│ Telegraf │───┼───▶│ InfluxDB │◀────Wire ─────────┘ -└─────────────┘ │ └───────────┘ Protocol - │ ▲ -┌─────────────┐ │ │ -│ HOST │ │ ▼ -│ Telegraf │───┘ ┌──────────┐ -└─────────────┘ │ Grafana │ - └──────────┘ -``` +## Description + +[Grafana](https://grafana.org/) is a graphing application built for time series data. It natively supports prometheus and provides great dashboarding support. This project is focused on provided a grafana installation that can be run within a kubernetes installation. The grafana application is agnostic to [Workflow][Workflow] and can be installed as a stand alone system with the monitoring suite. + +## Development + +The provided `Makefile` has various targets to help support building and publishing new images into a kubernetes cluster. + +### Environment variables + +There are a few key environment variables you should be aware of when interacting with the `make` targets. + +* `BUILD_TAG` - The tag provided to the podman image when it is built (defaults to the git-sha) +* `SHORT_NAME` - The name of the image (defaults to `grafana`) +* `DRYCC_REGISTRY` - This is the registry you are using (default `registry.drycc.cc`) +* `IMAGE_PREFIX` - This is the account for the registry you are using (default `drycc`) + +### Make targets + +* `make build` - Build container image +* `make push` - Push container image to a registry +* `make upgrade` - Replaces the running grafana instance with a new one + +The typical workflow will look something like this - `DRYCC_REGISTRY= IMAGE_PREFIX=foouser make build push upgrade` + [k8s-home]: http://kubernetes.io/ [issues]: https://github.com/drycc/monitor/issues [prs]: https://github.com/drycc/monitor/pulls -[v2.18]: https://github.com/drycc/workflow/releases/tag/v2.18.0 - +[workflow]: https://github.com/drycc/workflow \ No newline at end of file diff --git a/charts/monitor/Chart.yaml b/charts/grafana/Chart.yaml similarity index 57% rename from charts/monitor/Chart.yaml rename to charts/grafana/Chart.yaml index 5b237f6..f3bffea 100644 --- a/charts/monitor/Chart.yaml +++ b/charts/grafana/Chart.yaml @@ -1,11 +1,12 @@ -name: monitor -home: https://github.com/drycc/monitor +name: grafana +home: https://github.com/drycc/grafana apiVersion: v2 +appVersion: 1.0.0 dependencies: - name: common repository: oci://registry.drycc.cc/charts - version: ~1.1.1 -description: Monitoring for Drycc Workflow. + version: ~1.1.2 +description: Grafana for Drycc Workflow. maintainers: - name: Drycc Team email: engineering@drycc.com diff --git a/charts/grafana/templates/_helpers.tmpl b/charts/grafana/templates/_helpers.tmpl new file mode 100644 index 0000000..0519ef6 --- /dev/null +++ b/charts/grafana/templates/_helpers.tmpl @@ -0,0 +1,105 @@ +{{/* Generate grafana envs */}} +{{- define "grafana.envs" }} +env: +- name: POD_IP + valueFrom: + fieldRef: + fieldPath: status.podIP +- name: NAMESPACE + valueFrom: + fieldRef: + fieldPath: metadata.namespace +{{- if (.Values.valkeyUrl) }} +- name: DRYCC_VALKEY_URL + value: "{{ .Values.valkeyUrl }}" +{{- else if .Values.valkey.enabled }} +- name: DRYCC_VALKEY_PASSWORD + valueFrom: + secretKeyRef: + name: valkey-creds + key: password +- name: DRYCC_VALKEY_URL + value: "redis://:$(DRYCC_VALKEY_PASSWORD)@drycc-valkey:16379/2" +{{- end }} +- name: "DRYCC_CONTROLLER_URL" + value: http://drycc-controller-api +- name: "DRYCC_QUICKWIT_URL" + value: "$(DRYCC_CONTROLLER_URL)/v2/quickwit/drycc" +- name: "DRYCC_VICTORIAMETRICS_URL" + value: "$(DRYCC_CONTROLLER_URL)/v2/prometheus/drycc" +{{- if .Values.passport.enabled}} +- name: "DRYCC_PASSPORT_URL" +{{- if .Values.global.certManagerEnabled }} + value: https://drycc-passport.{{ .Values.global.platformDomain }} +{{- else }} + value: http://drycc-passport.{{ .Values.global.platformDomain }} +{{- end }} +- name: DRYCC_PASSPORT_KEY + valueFrom: + secretKeyRef: + name: passport-creds + key: drycc-passport-grafana-key +- name: DRYCC_PASSPORT_SECRET + valueFrom: + secretKeyRef: + name: passport-creds + key: drycc-passport-grafana-secret +- name: DRYCC_PASSPORT_SCOPES + valueFrom: + secretKeyRef: + name: passport-creds + key: drycc-passport-grafana-scopes +{{- else }} +- name: DRYCC_PASSPORT_URL + value: "{{ .Values.passportUrl }}" +- name: DRYCC_PASSPORT_KEY + value: "{{ .Values.passportKey }}" +- name: DRYCC_PASSPORT_SECRET + value: "{{ .Values.passportSecret }}" +- name: DRYCC_PASSPORT_SCOPES + value: "{{ .Values.passportScopes }}" +{{- end }} +- name: GF_DATABASE_TYPE + value: postgres +{{- if (.Values.databaseUrl) }} +- name: GF_DATABASE_URL + value: "{{ .Values.databaseUrl }}" +{{- else if .Values.database.enabled }} +- name: GF_DATABASE_USER + valueFrom: + secretKeyRef: + name: database-creds + key: user +- name: GF_DATABASE_PASSWORD + valueFrom: + secretKeyRef: + name: database-creds + key: password +- name: GF_DATABASE_URL + value: "postgres://$(GF_DATABASE_USER):$(GF_DATABASE_PASSWORD)@drycc-database:5432/grafana" +{{- end }} +- name: "GF_SECURITY_ADMIN_USER" + valueFrom: + secretKeyRef: + name: grafana-creds + key: admin-username +- name: "GF_SECURITY_ADMIN_PASSWORD" + valueFrom: + secretKeyRef: + name: grafana-creds + key: admin-password +- name: "GF_SERVER_ROOT_URL" +{{- if .Values.global.certManagerEnabled }} + value: https://drycc-grafana.{{ .Values.global.platformDomain }} +{{- else }} + value: http://drycc-grafana.{{ .Values.global.platformDomain }} +{{- end}} +- name: GF_UNIFIED_ALERTING_HA_PEERS + value: "drycc-grafana:9094" +- name: GF_UNIFIED_ALERTING_HA_ADVERTISE_ADDRESS + value: "$(POD_IP):9094" +{{- range $key, $value := .Values.environment }} +- name: {{ $key }} + value: {{ $value | quote }} +{{- end }} +{{- end }} diff --git a/charts/grafana/templates/grafana-configmap.yaml b/charts/grafana/templates/grafana-configmap.yaml new file mode 100644 index 0000000..0051a8b --- /dev/null +++ b/charts/grafana/templates/grafana-configmap.yaml @@ -0,0 +1,133 @@ +apiVersion: v1 +kind: ConfigMap +metadata: + name: grafana-config + namespace: {{ .Release.Namespace }} + labels: + heritage: drycc +data: + Caddyfile: | + { + order reverse_proxy before encode + auto_https off + log { + output stdout + } + } + :80 { + # For security reasons, deny external access to internal endpoints. + handle /proxy/* { + respond "Not Found" 404 + } + handle /alerts/* { + respond "Not Found" 404 + } + + handle /oauth2/* { + reverse_proxy 127.0.0.1:4000 + } + + handle { + forward_auth 127.0.0.1:4000 { + uri /oauth2/userinfo + copy_headers Remote-User Remote-Name Remote-Email + header_up -Connection + header_up X-Real-IP {remote_host} + @error status 401 + handle_response @error { + redir * /oauth2/sign_in?redirect={scheme}://{host}{uri} + } + } + reverse_proxy 127.0.0.1:3000 + } + } + grafana.ini: | + app_mode = production + + [paths] + data = /opt/drycc/grafana/data + home = /usr/share/grafana + logs = /opt/drycc/grafana/data/logs + plugins = /usr/share/grafana/plugins + provisioning = /usr/share/grafana/provisioning + + [server] + http_port = 3000 + router_logging = false + static_root_path = public + enable_gzip = true + + [analytics] + reporting_enabled = false + + [security] + admin_email = admin@drycc.cc + secret_key = KSJRZP0RIhO7P14e1TykTEzWuJhBobAj + login_remember_days = 7 + cookie_username = grafana_user + cookie_remember_name = grafana_remember + disable_gravatar = false + + [users] + allow_sign_up = false + allow_org_create = false + auto_assign_org = true + auto_assign_org_role = None + login_hint = email or username + + [auth] + disable_login_form = true + disable_signout_menu = false + + [auth.basic] + enabled = true + + [auth.proxy] + enabled = true + header_name = Remote-User + header_property = username + auto_sign_up = true + sync_ttl = 15 + headers = Name:Remote-Name Email:Remote-Email + enable_login_token = false + + [live] + max_connections = 100 + message_size_limit = 65536 + allowed_origins = * + ha_engine = redis + ha_prefix = "grafana:live" + + [emails] + welcome_email_on_sign_up = false + + [log] + mode = console + buffer_len = 10000 + level = info + + [log.console] + level = info + + [quota] + enabled = true + org_user = -1 + org_dashboard = 20 + org_data_source = 5 + org_api_key = 2 + org_alert_rule = 20 + user_org = -1 + alerting_rule_group_rules = 20 + alerting_rule_evaluation_results = 20 + + [unified_alerting] + enabled = true + ha_peer_timeout = 15s + ha_reconnect_timeout = 2m + ha_listen_address = 0.0.0.0:9094 + evaluation_timeout = 3s + max_attempts = 3 + min_interval = 1m + + [plugins] + plugin_admin_enabled = false diff --git a/charts/grafana/templates/grafana-cronjob-daily.yaml b/charts/grafana/templates/grafana-cronjob-daily.yaml new file mode 100644 index 0000000..8cd9a66 --- /dev/null +++ b/charts/grafana/templates/grafana-cronjob-daily.yaml @@ -0,0 +1,29 @@ +apiVersion: batch/v1 +kind: CronJob +metadata: + name: grafana-oauth2-token-refresher + labels: + app: grafana + component: token-refresher + heritage: drycc +spec: + schedule: "0 2 * * *" + concurrencyPolicy: Forbid + successfulJobsHistoryLimit: 3 + failedJobsHistoryLimit: 3 + jobTemplate: + spec: + backoffLimit: 3 + template: + metadata: + labels: + app: grafana + component: token-refresher + spec: + restartPolicy: OnFailure + containers: + - name: token-refresher + image: {{ .Values.imageRegistry }}/{{ .Values.imageOrg }}/grafana:{{ .Values.imageTag }} + imagePullPolicy: {{ .Values.imagePullPolicy }} + command: ["/usr/bin/env", "python3", "/usr/share/grafana/oauth2/credentials.py"] + {{- include "grafana.envs" . | indent 12 }} diff --git a/charts/grafana/templates/grafana-secret-creds.yaml b/charts/grafana/templates/grafana-secret-creds.yaml new file mode 100644 index 0000000..1ea2b53 --- /dev/null +++ b/charts/grafana/templates/grafana-secret-creds.yaml @@ -0,0 +1,10 @@ +apiVersion: v1 +kind: Secret +metadata: + name: grafana-creds + labels: + heritage: drycc +type: Opaque +data: + admin-username: {{ include "common.secrets.lookup" (dict "secret" "grafana-creds" "key" "admin-username" "defaultValue" (default (randAlphaNum 16) .Values.adminUsername) "context" $) }} + admin-password: {{ include "common.secrets.lookup" (dict "secret" "grafana-creds" "key" "admin-password" "defaultValue" (default (randAlphaNum 32) .Values.adminPassword) "context" $) }} diff --git a/charts/grafana/templates/grafana-service.yaml b/charts/grafana/templates/grafana-service.yaml new file mode 100644 index 0000000..b570e7a --- /dev/null +++ b/charts/grafana/templates/grafana-service.yaml @@ -0,0 +1,33 @@ +apiVersion: v1 +kind: Service +metadata: + name: drycc-grafana + annotations: + {{- with .Values.service.annotations }} + {{- toYaml . | nindent 4 }} + {{- end }} + labels: + heritage: drycc + app: drycc-grafana +spec: + type: ClusterIP + publishNotReadyAddresses: true + ports: + - name: grafana-tcp + port: 3000 + targetPort: grafana + protocol: TCP + - name: alerting-tcp + port: 9094 + targetPort: alerting + protocol: TCP + - name: alerting-udp + port: 9094 + targetPort: alerting + protocol: UDP + - name: proxy-tcp + port: 80 + targetPort: proxy + protocol: TCP + selector: + app: drycc-grafana diff --git a/charts/grafana/templates/grafana-statefulset.yaml b/charts/grafana/templates/grafana-statefulset.yaml new file mode 100644 index 0000000..1696c9a --- /dev/null +++ b/charts/grafana/templates/grafana-statefulset.yaml @@ -0,0 +1,293 @@ +apiVersion: apps/v1 +kind: StatefulSet +metadata: + name: drycc-grafana + labels: + heritage: drycc + annotations: + component.drycc.cc/version: {{ .Values.imageTag }} +spec: + replicas: 1 + updateStrategy: + type: RollingUpdate + podManagementPolicy: Parallel + selector: + matchLabels: + app: drycc-grafana + template: + metadata: + labels: {{- include "common.labels.standard" . | nindent 8 }} + app: drycc-grafana + spec: + affinity: + podAffinity: {{- include "common.affinities.pods" (dict "type" .Values.podAffinityPreset.type "component" "" "extraMatchLabels" .Values.podAffinityPreset.extraMatchLabels "topologyKey" "" "context" $) | nindent 10 }} + podAntiAffinity: {{- include "common.affinities.pods" (dict "type" .Values.podAntiAffinityPreset.type "component" "" "extraMatchLabels" .Values.podAntiAffinityPreset.extraMatchLabels "topologyKey" "" "context" $) | nindent 10 }} + nodeAffinity: {{- include "common.affinities.nodes" (dict "type" .Values.nodeAffinityPreset.type "key" .Values.nodeAffinityPreset.key "values" .Values.nodeAffinityPreset.values ) | nindent 10 }} + securityContext: + fsGroup: 1001 + runAsGroup: 1001 + runAsUser: 1001 + initContainers: + - name: drycc-grafana-init + image: {{.Values.imageRegistry}}/{{.Values.imageOrg}}/python-dev:latest + imagePullPolicy: {{.Values.imagePullPolicy}} + args: + - netcat + - -v + - -u + - $(GF_DATABASE_URL),http://drycc-controller-api + {{- include "grafana.envs" . | indent 8 }} + - name: drycc-grafana-migrate + image: {{.Values.imageRegistry}}/{{.Values.imageOrg}}/grafana:{{.Values.imageTag}} + imagePullPolicy: {{.Values.imagePullPolicy}} + args: + - /usr/bin/env + - bash + - -ec + - | + set -m + echo "Configure Grafana live HA engine" + DRYCC_VALKEY_JSON=$(echo "$DRYCC_VALKEY_URL" | jq -R 'capture("^(?redis[s]?)://(?[^:]+)?(?::(?[^@]+))?@(?
[^/]+)/?(?[0-9]*)$")') + GF_LIVE_HA_ENGINE_ADDRESS=$(echo "${DRYCC_VALKEY_JSON}" |jq -r '.address') + GF_LIVE_HA_ENGINE_PASSWORD=$(echo "${DRYCC_VALKEY_JSON}" |jq -r '.password') + export GF_LIVE_HA_ENGINE_ADDRESS GF_LIVE_HA_ENGINE_PASSWORD + grafana server --config /usr/share/grafana/grafana.ini --homepath /opt/drycc/grafana & + GRAFANA_PID=$! + echo "Waiting for Grafana to come up..." + until curl -q --fail --output /dev/null --silent "http://localhost:3000/api/health"; do + printf "." + sleep 2 + done + kill $GRAFANA_PID 2>/dev/null + wait $GRAFANA_PID 2>/dev/null + {{- include "grafana.envs" . | indent 8 }} + volumeMounts: + {{- if .Values.persistence.enabled }} + - name: grafana-data + mountPath: /var/lib/grafana + {{- end }} + - name: grafana-config + subPath: grafana.ini + readOnly: true + mountPath: /usr/share/grafana/grafana.ini + containers: + - name: drycc-grafana + image: {{.Values.imageRegistry}}/{{.Values.imageOrg}}/grafana:{{.Values.imageTag}} + imagePullPolicy: {{.Values.imagePullPolicy}} + {{- if .Values.diagnosticMode.enabled }} + command: {{- include "common.tplvalues.render" (dict "value" .Values.diagnosticMode.command "context" $) | nindent 10 }} + args: {{- include "common.tplvalues.render" (dict "value" .Values.diagnosticMode.args "context" $) | nindent 10 }} + {{- else }} + args: + - /usr/bin/env + - bash + - -ec + - | + set -m + echo "Configure Grafana live HA engine" + DRYCC_VALKEY_JSON=$(echo "$DRYCC_VALKEY_URL" | jq -R 'capture("^(?redis[s]?)://(?[^:]+)?(?::(?[^@]+))?@(?
[^/]+)/?(?[0-9]*)$")') + GF_LIVE_HA_ENGINE_ADDRESS=$(echo "${DRYCC_VALKEY_JSON}" |jq -r '.address') + GF_LIVE_HA_ENGINE_PASSWORD=$(echo "${DRYCC_VALKEY_JSON}" |jq -r '.password') + export GF_LIVE_HA_ENGINE_ADDRESS GF_LIVE_HA_ENGINE_PASSWORD + exec grafana server --config /usr/share/grafana/grafana.ini --homepath /opt/drycc/grafana + {{- end }} + {{- with index .Values "resources" }} + resources: + {{- toYaml . | nindent 10 }} + {{- end }} + ports: + - containerPort: 3000 + name: grafana + - containerPort: 9094 + name: alerting + {{- include "grafana.envs" . | indent 8 }} + {{- if not .Values.diagnosticMode.enabled }} + livenessProbe: + tcpSocket: + port: grafana + initialDelaySeconds: 120 + periodSeconds: 10 + timeoutSeconds: 5 + failureThreshold: 6 + successThreshold: 1 + readinessProbe: + httpGet: + path: /api/health + port: grafana + initialDelaySeconds: 30 + periodSeconds: 10 + timeoutSeconds: 5 + failureThreshold: 6 + successThreshold: 1 + startupProbe: + httpGet: + path: /api/health + port: grafana + initialDelaySeconds: 120 + periodSeconds: 10 + timeoutSeconds: 5 + failureThreshold: 6 + successThreshold: 1 + {{- end }} + volumeMounts: + {{- if .Values.persistence.enabled }} + - name: grafana-data + mountPath: /var/lib/grafana + {{- end }} + - name: grafana-config + subPath: grafana.ini + readOnly: true + mountPath: /usr/share/grafana/grafana.ini + - name: drycc-grafana-oauth2 + image: {{.Values.imageRegistry}}/{{.Values.imageOrg}}/grafana:{{.Values.imageTag}} + imagePullPolicy: {{.Values.imagePullPolicy}} + {{- if .Values.diagnosticMode.enabled }} + command: {{- include "common.tplvalues.render" (dict "value" .Values.diagnosticMode.command "context" $) | nindent 10 }} + args: {{- include "common.tplvalues.render" (dict "value" .Values.diagnosticMode.args "context" $) | nindent 10 }} + {{- else }} + args: + - /usr/bin/env + - bash + - -ec + - | + set -m + echo "Waiting for Grafana to come up..." + until curl -q --fail --output /dev/null --silent "http://${GF_SECURITY_ADMIN_USER}:${GF_SECURITY_ADMIN_PASSWORD}@localhost:${GF_SERVER_HTTP_PORT:-3000}/api/org"; do + printf "." + sleep 2 + done + echo "Grafana is up and running." + python3 oauth2/main.py \ + --bind 127.0.0.1 \ + --port 4000 \ + --client-id $(DRYCC_PASSPORT_KEY) \ + --client-secret $(DRYCC_PASSPORT_SECRET) \ + --oidc-issuer-url $(DRYCC_PASSPORT_URL)/oauth + {{- end }} + {{- with index .Values "resources" }} + resources: + {{- toYaml . | nindent 10 }} + {{- end }} + {{- include "grafana.envs" . | indent 8 }} + {{- if not .Values.diagnosticMode.enabled }} + livenessProbe: + exec: + command: + - curl + - -fsS + - -o + - /dev/null + - http://127.0.0.1:4000/oauth2/healthz + initialDelaySeconds: 240 + periodSeconds: 10 + timeoutSeconds: 5 + failureThreshold: 6 + successThreshold: 1 + readinessProbe: + exec: + command: + - curl + - -fsS + - -o + - /dev/null + - http://127.0.0.1:4000/oauth2/healthz + initialDelaySeconds: 30 + periodSeconds: 10 + timeoutSeconds: 5 + failureThreshold: 6 + successThreshold: 1 + startupProbe: + exec: + command: + - curl + - -fsS + - -o + - /dev/null + - http://127.0.0.1:4000/oauth2/healthz + initialDelaySeconds: 30 + periodSeconds: 10 + timeoutSeconds: 5 + failureThreshold: 6 + successThreshold: 1 + {{- end }} + - name: drycc-grafana-proxy + image: {{.Values.imageRegistry}}/{{.Values.imageOrg}}/grafana:{{.Values.imageTag}} + imagePullPolicy: {{.Values.imagePullPolicy}} + {{- if .Values.diagnosticMode.enabled }} + command: {{- include "common.tplvalues.render" (dict "value" .Values.diagnosticMode.command "context" $) | nindent 10 }} + args: {{- include "common.tplvalues.render" (dict "value" .Values.diagnosticMode.args "context" $) | nindent 10 }} + {{- else }} + args: + - /usr/bin/env + - bash + - -ec + - | + set -m + echo "Waiting for Grafana oauth2 proxy to come up..." + until curl -q --fail --output /dev/null --silent "http://localhost:4000/oauth2/healthz"; do + printf "." + sleep 2 + done + echo "Grafana oauth2 proxy is up and running." + caddy run --config /usr/share/grafana/Caddyfile + {{- end }} + {{- with index .Values "resources" }} + resources: + {{- toYaml . | nindent 10 }} + {{- end }} + ports: + - containerPort: 80 + name: proxy + {{- include "grafana.envs" . | indent 8 }} + {{- if not .Values.diagnosticMode.enabled }} + livenessProbe: + tcpSocket: + port: proxy + initialDelaySeconds: 300 + periodSeconds: 10 + timeoutSeconds: 5 + failureThreshold: 6 + successThreshold: 1 + readinessProbe: + httpGet: + path: /oauth2/healthz + port: proxy + initialDelaySeconds: 30 + periodSeconds: 10 + timeoutSeconds: 5 + failureThreshold: 6 + successThreshold: 1 + startupProbe: + httpGet: + path: /oauth2/healthz + port: proxy + initialDelaySeconds: 30 + periodSeconds: 10 + timeoutSeconds: 5 + failureThreshold: 6 + successThreshold: 1 + {{- end }} + volumeMounts: + {{- if .Values.persistence.enabled }} + - name: grafana-data + mountPath: /opt/drycc/grafana/data + {{- end }} + - name: grafana-config + subPath: Caddyfile + readOnly: true + mountPath: /usr/share/grafana/Caddyfile + volumes: + - name: grafana-config + configMap: + name: grafana-config + {{- if .Values.persistence.enabled }} + volumeClaimTemplates: + - metadata: + name: grafana-data + spec: + accessModes: [ "ReadWriteOnce" ] + storageClassName: {{ .Values.persistence.storageClass }} + resources: + requests: + storage: {{ .Values.persistence.size | quote }} + {{- end }} diff --git a/charts/grafana/values.yaml b/charts/grafana/values.yaml new file mode 100644 index 0000000..9550866 --- /dev/null +++ b/charts/grafana/values.yaml @@ -0,0 +1,104 @@ +imageOrg: "drycc" +imagePullPolicy: "Always" +imageTag: "canary" +imageRegistry: "registry.drycc.cc" + +## Enable diagnostic mode +## +diagnosticMode: + ## @param diagnosticMode.enabled Enable diagnostic mode (all probes will be disabled and the command will be overridden) + ## + enabled: false + ## @param diagnosticMode.command Command to override all containers + ## + command: + - sleep + ## @param diagnosticMode.args Args to override all containers + ## + args: + - infinity + +resources: {} + # limits: + # cpu: 200m + # memory: 50Mi + # requests: + # cpu: 100m + # memory: 30Mi + +nodeAffinityPreset: + key: "drycc.cc/node" + type: "soft" + values: + - "true" + +podAffinityPreset: + type: "" + extraMatchLabels: + security: "drycc-security" + +podAntiAffinityPreset: + type: "soft" + extraMatchLabels: + app: "drycc-grafana" + +# grafana security config +adminUsername: "" +adminPassword: "" + +# Service +service: + # Provide any additional service annotations + annotations: {} + +persistence: + enabled: false + size: 5Gi + storageClass: "" +# Any custom grafana environment variables +# can be specified as key-value pairs under environment +# this is usually a non required setting. +environment: {} + +## valkeyUrl are will no longer use the built-in valkey component +valkeyUrl: "" +## databaseUrl are will no longer use the built-in database component +databaseUrl: "" +# The following parameters are used when passport.enabled is false +passportUrl: "" +passportKey: "" +passportSecret: "" +passportScopes: "" +# victoriametricsUrl is will no longer use the built-in victoriametrics component +victoriametricsUrl: "" + +valkey: + enabled: true + replicas: 1 + port: "6379" + +grafana: + enabled: true + +storage: + enabled: true + +database: + enabled: true + +registry: + enabled: true + +passport: + enabled: true + +victoriametrics: + enabled: true + +global: + # The public resolvable hostname to build your cluster with. + # + # This will be the hostname that is used to build endpoints such as "drycc.$HOSTNAME" + platformDomain: "" + # Whether cert_manager is enabled to automatically generate drycc certificates + certManagerEnabled: true diff --git a/charts/monitor/templates/_helpers.tmpl b/charts/monitor/templates/_helpers.tmpl deleted file mode 100644 index ec93696..0000000 --- a/charts/monitor/templates/_helpers.tmpl +++ /dev/null @@ -1,67 +0,0 @@ -{{/* Generate monitor grafana envs */}} -{{- define "grafana.envs" }} -env: -{{- if eq .Values.global.influxdbLocation "off-cluster" }} -- name: "INFLUXDB_URL" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: url -{{- else }} -- name: "INFLUXDB_URL" - value: http://$(DRYCC_INFLUXDB_SERVICE_HOST):$(DRYCC_INFLUXDB_SERVICE_PORT) -{{- end }} -- name: "INFLUXDB_BUCKET" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: bucket -- name: "INFLUXDB_ORG" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: org -- name: "INFLUXDB_TOKEN" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: token -- name: "BIND_PORT" - value: "3000" -{{- if not (.Values.grafana.environment.DEFAULT_USER) }} -- name: "DEFAULT_USER" - value: {{ randAlphaNum 32 | b64enc }} -{{- end}} -{{- if not (.Values.grafana.environment.DEFAULT_USER_PASSWORD) }} -- name: "DEFAULT_USER_PASSWORD" - value: {{ randAlphaNum 32 | b64enc }} -{{- end}} -- name: "KUBERNETES_CLUSTER_DOMAIN" - value: {{.Values.global.clusterDomain}} -{{- range $key, $value := .Values.grafana.environment }} -- name: {{ $key }} - value: {{ $value | quote }} -{{- end }} -- name: "DRYCC_GRAFANA_URL" -{{- if .Values.global.certManagerEnabled }} - value: https://drycc-monitor-grafana.{{ .Values.global.platformDomain }} -{{- else }} - value: http://drycc-monitor-grafana.{{ .Values.global.platformDomain }} -{{- end }} -- name: "DRYCC_PASSPORT_URL" -{{- if .Values.global.certManagerEnabled }} - value: https://drycc-passport.{{ .Values.global.platformDomain }} -{{- else }} - value: http://drycc-passport.{{ .Values.global.platformDomain }} -{{- end }} -- name: DRYCC_PASSPORT_KEY - valueFrom: - secretKeyRef: - name: passport-creds - key: drycc-passport-grafana-key -- name: DRYCC_PASSPORT_SECRET - valueFrom: - secretKeyRef: - name: passport-creds - key: drycc-passport-grafana-secret -{{- end }} diff --git a/charts/monitor/templates/monitor-clusterrole.yaml b/charts/monitor/templates/monitor-clusterrole.yaml deleted file mode 100644 index 2a4ed33..0000000 --- a/charts/monitor/templates/monitor-clusterrole.yaml +++ /dev/null @@ -1,14 +0,0 @@ -{{- if (.Values.global.rbac) -}} -# Define global role with the default system:aggregate-to-view cluster role and the two rules we just created -kind: ClusterRole -apiVersion: rbac.authorization.k8s.io/v1 -metadata: - name: drycc:drycc-monitor -aggregationRule: - clusterRoleSelectors: - - matchLabels: - rbac.authorization.k8s.io/aggregate-view-telegraf-stats: "true" - - matchLabels: - rbac.authorization.k8s.io/aggregate-to-view: "true" -rules: [] # Rules are automatically filled in by the controller manager. -{{- end -}} diff --git a/charts/monitor/templates/monitor-grafana-certificate.yaml b/charts/monitor/templates/monitor-grafana-certificate.yaml deleted file mode 100644 index 99fb705..0000000 --- a/charts/monitor/templates/monitor-grafana-certificate.yaml +++ /dev/null @@ -1,15 +0,0 @@ -{{ if .Values.global.certManagerEnabled }} -apiVersion: cert-manager.io/v1 -kind: Certificate -metadata: - name: drycc-monitor-grafana -spec: - secretName: drycc-monitor-grafana-auto-tls - issuerRef: - name: drycc-cluster-issuer - kind: ClusterIssuer - dnsNames: - - drycc-monitor-grafana.{{ .Values.global.platformDomain }} - privateKey: - rotationPolicy: Always -{{- end }} diff --git a/charts/monitor/templates/monitor-grafana-deployment.yaml b/charts/monitor/templates/monitor-grafana-deployment.yaml deleted file mode 100644 index 8ab955b..0000000 --- a/charts/monitor/templates/monitor-grafana-deployment.yaml +++ /dev/null @@ -1,60 +0,0 @@ -{{- if eq .Values.global.grafanaLocation "on-cluster" }} -apiVersion: apps/v1 -kind: Deployment -metadata: - name: drycc-monitor-grafana - labels: - heritage: drycc - annotations: - component.drycc.cc/version: {{ .Values.grafana.imageTag }} -spec: - replicas: 1 - strategy: - rollingUpdate: - maxSurge: 1 - maxUnavailable: 0 - type: RollingUpdate - selector: - matchLabels: - app: drycc-monitor-grafana - template: - metadata: - labels: - app: drycc-monitor-grafana - spec: - affinity: - podAffinity: {{- include "common.affinities.pods" (dict "type" .Values.grafana.podAffinityPreset.type "key" .Values.grafana.podAffinityPreset.key "values" .Values.grafana.podAffinityPreset.values ) | nindent 10 }} - podAntiAffinity: {{- include "common.affinities.pods" (dict "type" .Values.grafana.podAntiAffinityPreset.type "key" .Values.grafana.podAntiAffinityPreset.key "values" .Values.grafana.podAntiAffinityPreset.values ) | nindent 10 }} - nodeAffinity: {{- include "common.affinities.nodes" (dict "type" .Values.grafana.nodeAffinityPreset.type "key" .Values.grafana.nodeAffinityPreset.key "values" .Values.grafana.nodeAffinityPreset.values ) | nindent 10 }} - securityContext: - fsGroup: 1001 - runAsGroup: 1001 - runAsUser: 1001 - containers: - - name: drycc-monitor-grafana - image: {{.Values.grafana.imageRegistry}}/{{.Values.grafana.imageOrg}}/grafana:{{.Values.grafana.imageTag}} - imagePullPolicy: {{.Values.grafana.imagePullPolicy}} - {{- if or (.Values.grafana.limitsCpu) (.Values.grafana.limitsMemory)}} - resources: - limits: - {{- if (.Values.grafana.limitsCpu) }} - cpu: {{.Values.grafana.limitsCpu}} - {{- end}} - {{- if (.Values.grafana.limitsMemory) }} - memory: {{.Values.grafana.limitsMemory}} - {{- end}} - {{- end}} - ports: - - containerPort: 3000 - name: ui - {{- include "grafana.envs" . | indent 8 }} - {{- if .Values.grafana.persistence.enabled }} - volumeMounts: - - name: grafana-data - mountPath: /var/lib/grafana - volumes: - - name: grafana-data - persistentVolumeClaim: - claimName: drycc-monitor-grafana - {{- end }} -{{- end }} diff --git a/charts/monitor/templates/monitor-grafana-ingress.yaml b/charts/monitor/templates/monitor-grafana-ingress.yaml deleted file mode 100644 index 0d88091..0000000 --- a/charts/monitor/templates/monitor-grafana-ingress.yaml +++ /dev/null @@ -1,36 +0,0 @@ -apiVersion: networking.k8s.io/v1 -kind: Ingress -metadata: - name: "drycc-monitor-grafana" - labels: - app: "drycc-monitor-grafana" - chart: "{{ .Chart.Name }}-{{ .Chart.Version }}" - release: "{{ .Release.Name }}" - heritage: "{{ .Release.Service }}" - annotations: - kubernetes.io/tls-acme: "true" -spec: - {{- if not (eq .Values.global.ingressClass "") }} - ingressClassName: "{{ .Values.global.ingressClass }}" - {{ end }} - rules: - - host: drycc-monitor-grafana.{{ .Values.global.platformDomain }} - http: - paths: - - pathType: Prefix - {{- if eq .Values.global.ingressClass "gce" "alb" }} - path: /* - {{- else }}{{/* Has annotations but ingress class is not "gce" nor "alb" */}} - path: / - {{- end }} - backend: - service: - name: drycc-monitor-grafana - port: - number: 80 - {{ if .Values.global.certManagerEnabled }} - tls: - - secretName: drycc-monitor-grafana-auto-tls - hosts: - - drycc-monitor-grafana.{{ .Values.global.platformDomain }} - {{- end }} diff --git a/charts/monitor/templates/monitor-grafana-pvc.yaml b/charts/monitor/templates/monitor-grafana-pvc.yaml deleted file mode 100644 index ea1059a..0000000 --- a/charts/monitor/templates/monitor-grafana-pvc.yaml +++ /dev/null @@ -1,24 +0,0 @@ -{{- if .Values.grafana.persistence.enabled }} -kind: PersistentVolumeClaim -apiVersion: v1 -metadata: - name: drycc-monitor-grafana - labels: - heritage: drycc -spec: - accessModes: - - {{ .Values.grafana.persistence.accessMode | quote }} - resources: - requests: - storage: {{ .Values.grafana.persistence.size | quote }} -{{- if .Values.grafana.persistence.storageClass }} -{{- if (eq "-" .Values.grafana.persistence.storageClass) }} - storageClassName: "" -{{- else }} - storageClassName: "{{ .Values.grafana.persistence.storageClass }}" -{{- end }} -{{- end }} -{{- if .Values.grafana.persistence.volumeName }} - volumeName: "{{ .Values.grafana.persistence.volumeName }}" -{{- end }} -{{- end }} diff --git a/charts/monitor/templates/monitor-grafana-svc.yaml b/charts/monitor/templates/monitor-grafana-svc.yaml deleted file mode 100644 index 23cf24d..0000000 --- a/charts/monitor/templates/monitor-grafana-svc.yaml +++ /dev/null @@ -1,20 +0,0 @@ -{{- if eq .Values.global.grafanaLocation "on-cluster" }} -apiVersion: v1 -kind: Service -metadata: - name: drycc-monitor-grafana - annotations: - {{- with .Values.grafana.service.annotations }} - {{- toYaml . | nindent 4 }} - {{- end }} - labels: - heritage: drycc - app: drycc-monitor-grafana -spec: - ports: - - port: 80 - name: ui - targetPort: ui - selector: - app: drycc-monitor-grafana -{{- end }} diff --git a/charts/monitor/templates/monitor-telegraf-clusterrole.yaml b/charts/monitor/templates/monitor-telegraf-clusterrole.yaml deleted file mode 100644 index 7e1c210..0000000 --- a/charts/monitor/templates/monitor-telegraf-clusterrole.yaml +++ /dev/null @@ -1,17 +0,0 @@ -{{- if (.Values.global.rbac) -}} -kind: ClusterRole -apiVersion: rbac.authorization.k8s.io/v1 -metadata: - name: drycc:drycc-monitor-telegraf - labels: - rbac.authorization.k8s.io/aggregate-view-telegraf-stats: "true" - app: drycc-monitor-telegraf - heritage: drycc -rules: - - apiGroups: ["metrics.k8s.io"] - resources: ["pods"] - verbs: ["get", "list", "watch"] - - apiGroups: [""] - resources: ["pods", "persistentvolumes", "nodes", "nodes/proxy", "nodes/stats"] - verbs: ["get", "list", "watch"] -{{- end -}} diff --git a/charts/monitor/templates/monitor-telegraf-clusterrolebinding.yaml b/charts/monitor/templates/monitor-telegraf-clusterrolebinding.yaml deleted file mode 100644 index 013a2ed..0000000 --- a/charts/monitor/templates/monitor-telegraf-clusterrolebinding.yaml +++ /dev/null @@ -1,18 +0,0 @@ -{{- if (.Values.global.rbac) -}} -kind: ClusterRoleBinding -apiVersion: rbac.authorization.k8s.io/v1 -metadata: - name: drycc:drycc-monitor-telegraf - labels: - app: drycc-monitor-telegraf - heritage: drycc -roleRef: - apiGroup: rbac.authorization.k8s.io - kind: ClusterRole - name: drycc:drycc-monitor-telegraf -subjects: -- kind: ServiceAccount - name: drycc-monitor-telegraf - namespace: {{ .Release.Namespace }} -{{- end -}} - diff --git a/charts/monitor/templates/monitor-telegraf-daemon.yaml b/charts/monitor/templates/monitor-telegraf-daemon.yaml deleted file mode 100644 index dc16f3c..0000000 --- a/charts/monitor/templates/monitor-telegraf-daemon.yaml +++ /dev/null @@ -1,135 +0,0 @@ -apiVersion: apps/v1 -kind: DaemonSet -metadata: - name: drycc-monitor-telegraf - labels: - heritage: drycc - annotations: - component.drycc.cc/version: {{ .Values.telegraf.imageTag }} -spec: - selector: - matchLabels: - app: drycc-monitor-telegraf - {{- if ge .Capabilities.KubeVersion.Minor "6" }} - updateStrategy: - type: RollingUpdate - {{- end }} - template: - metadata: - labels: - app: drycc-monitor-telegraf - spec: - securityContext: - fsGroup: 1001 - runAsGroup: 1001 - runAsUser: 1001 - serviceAccount: drycc-monitor-telegraf - initContainers: - - name: drycc-monitor-telegraf-init - image: {{.Values.telegraf.imageRegistry}}/{{.Values.telegraf.imageOrg}}/python-dev:latest - imagePullPolicy: {{.Values.imagePullPolicy}} - args: - - netcat - - -v - - -u - - $(INFLUXDB_URLS_V2) - env: -{{- if eq .Values.global.influxdbLocation "off-cluster" }} - - name: "INFLUXDB_URLS_V2" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: url -{{- else }} - - name: "INFLUXDB_URLS_V2" - value: "\"http://$(DRYCC_INFLUXDB_SERVICE_HOST):$(DRYCC_INFLUXDB_SERVICE_PORT)\"" - - name: "INFLUXDB_V2_INPUT_URLS" - value: "\"http://$(DRYCC_INFLUXDB_SERVICE_HOST):$(DRYCC_INFLUXDB_SERVICE_PORT)/metrics\"" - - name: "ENABLE_INFLUXDB_V2_INPUT" - value: "true" -{{- end }} - containers: - - name: drycc-monitor-telegraf - image: {{.Values.telegraf.imageRegistry}}/{{.Values.telegraf.imageOrg}}/telegraf:{{.Values.telegraf.imageTag}} - imagePullPolicy: {{.Values.telegraf.imagePullPolicy}} -{{- if or (.Values.telegraf.limitsCpu) (.Values.telegraf.limitsMemory)}} - resources: - limits: -{{- if (.Values.telegraf.limitsCpu) }} - cpu: {{.Values.telegraf.limitsCpu}} -{{- end}} -{{- if (.Values.telegraf.limitsMemory) }} - memory: {{.Values.telegraf.limitsMemory}} -{{- end}} -{{- end}} - env: - - name: POD_NAMESPACE - valueFrom: - fieldRef: - fieldPath: metadata.namespace -{{- if eq .Values.global.influxdbLocation "off-cluster" }} - - name: "INFLUXDB_URLS_V2" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: url -{{- else }} - - name: "INFLUXDB_URLS_V2" - value: "\"http://$(DRYCC_INFLUXDB_SERVICE_HOST):$(DRYCC_INFLUXDB_SERVICE_PORT)\"" - - name: "INFLUXDB_V2_INPUT_URLS" - value: "\"http://$(DRYCC_INFLUXDB_SERVICE_HOST):$(DRYCC_INFLUXDB_SERVICE_PORT)/metrics\"" - - name: "ENABLE_INFLUXDB_V2_INPUT" - value: "true" -{{- end }} - - name: "INFLUXDB_BUCKET" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: bucket - - name: "INFLUXDB_ORG" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: org - - name: "INFLUXDB_TOKEN" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: token - - name: "HOST_PROC" - value: "/rootfs/proc" - - name: "HOST_SYS" - value: "/rootfs/sys" - - name: "AGENT_QUIET" - value: "true" - - name: "AGENT_BUFFER_LIMIT" - value: "100000" - - name: "ENABLE_KUBERNETES" - value: "true" - - name: "ENABLE_BARE_METAL_SERVER" - value: "true" - volumeMounts: - - mountPath: /var/run/docker.sock - name: docker-socket - - name: sysro - mountPath: /rootfs/sys - readOnly: true - - name: procro - mountPath: /rootfs/proc - readOnly: true - - name: varrunutmpro - mountPath: /var/run/utmp - readOnly: true - volumes: - - name: docker-socket - hostPath: - path: /var/run/docker.sock - - name: sysro - hostPath: - path: /sys - - name: procro - hostPath: - path: /proc - - name: varrunutmpro - hostPath: - path: /var/run/utmp diff --git a/charts/monitor/templates/monitor-telegraf-deployment.yaml b/charts/monitor/templates/monitor-telegraf-deployment.yaml deleted file mode 100644 index 88da852..0000000 --- a/charts/monitor/templates/monitor-telegraf-deployment.yaml +++ /dev/null @@ -1,119 +0,0 @@ -apiVersion: apps/v1 -kind: Deployment -metadata: - name: drycc-monitor-telegraf - labels: - heritage: drycc - annotations: - component.drycc.cc/version: {{ .Values.telegraf.imageTag }} -spec: - replicas: 1 - selector: - matchLabels: - app: drycc-monitor-telegraf - {{- if ge .Capabilities.KubeVersion.Minor "6" }} - updateStrategy: - type: RollingUpdate - {{- end }} - template: - metadata: - labels: - app: drycc-monitor-telegraf - spec: - affinity: - podAffinity: {{- include "common.affinities.pods" (dict "type" .Values.telegraf.podAffinityPreset.type "key" .Values.telegraf.podAffinityPreset.key "values" .Values.telegraf.podAffinityPreset.values ) | nindent 10 }} - podAntiAffinity: {{- include "common.affinities.pods" (dict "type" .Values.telegraf.podAntiAffinityPreset.type "key" .Values.telegraf.podAntiAffinityPreset.key "values" .Values.telegraf.podAntiAffinityPreset.values ) | nindent 10 }} - nodeAffinity: {{- include "common.affinities.nodes" (dict "type" .Values.telegraf.nodeAffinityPreset.type "key" .Values.telegraf.nodeAffinityPreset.key "values" .Values.telegraf.nodeAffinityPreset.values ) | nindent 10 }} - securityContext: - fsGroup: 1001 - runAsGroup: 1001 - runAsUser: 1001 - serviceAccount: drycc-monitor-telegraf - initContainers: - - name: drycc-monitor-telegraf-init - image: {{.Values.telegraf.imageRegistry}}/{{.Values.telegraf.imageOrg}}/python-dev:latest - imagePullPolicy: {{.Values.imagePullPolicy}} - args: - - netcat - - -v - - -u - - $(INFLUXDB_URLS_V2) - env: -{{- if eq .Values.global.influxdbLocation "off-cluster" }} - - name: "INFLUXDB_URLS_V2" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: url -{{- else }} - - name: "INFLUXDB_URLS_V2" - value: "\"http://$(DRYCC_INFLUXDB_SERVICE_HOST):$(DRYCC_INFLUXDB_SERVICE_PORT)\"" - - name: "INFLUXDB_V2_INPUT_URLS" - value: "\"http://$(DRYCC_INFLUXDB_SERVICE_HOST):$(DRYCC_INFLUXDB_SERVICE_PORT)/metrics\"" - - name: "ENABLE_INFLUXDB_V2_INPUT" - value: "true" -{{- end }} - containers: - - name: drycc-monitor-telegraf - image: {{.Values.telegraf.imageRegistry}}/{{.Values.telegraf.imageOrg}}/telegraf:{{.Values.telegraf.imageTag}} - imagePullPolicy: {{.Values.telegraf.imagePullPolicy}} -{{- if or (.Values.telegraf.limitsCpu) (.Values.telegraf.limitsMemory)}} - resources: - limits: -{{- if (.Values.telegraf.limitsCpu) }} - cpu: {{.Values.telegraf.limitsCpu}} -{{- end}} -{{- if (.Values.telegraf.limitsMemory) }} - memory: {{.Values.telegraf.limitsMemory}} -{{- end}} -{{- end}} - env: - - name: POD_NAMESPACE - valueFrom: - fieldRef: - fieldPath: metadata.namespace - {{- if eq .Values.global.influxdbLocation "off-cluster" }} - - name: "INFLUXDB_URLS_V2" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: url - {{- else }} - - name: "INFLUXDB_URLS_V2" - value: "\"http://$(DRYCC_INFLUXDB_SERVICE_HOST):$(DRYCC_INFLUXDB_SERVICE_PORT)\"" - - name: "INFLUXDB_V2_INPUT_URLS" - value: "\"http://$(DRYCC_INFLUXDB_SERVICE_HOST):$(DRYCC_INFLUXDB_SERVICE_PORT)/metrics\"" - - name: "ENABLE_INFLUXDB_V2_INPUT" - value: "true" - {{- end }} - - name: "INFLUXDB_BUCKET" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: bucket - - name: "INFLUXDB_ORG" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: org - - name: "INFLUXDB_TOKEN" - valueFrom: - secretKeyRef: - name: influxdb-creds - key: token - - name: "AGENT_QUIET" - value: "true" - - name: "AGENT_BUFFER_LIMIT" - value: "100000" -{{- if eq .Values.global.redisLocation "on-cluster" }} - - name: DRYCC_REDIS_ADDRS - valueFrom: - secretKeyRef: - name: redis-creds - key: addrs - - name: DRYCC_REDIS_PASSWORD - valueFrom: - secretKeyRef: - name: redis-creds - key: password -{{- end }} diff --git a/charts/monitor/templates/monitor-telegraf-service-account.yaml b/charts/monitor/templates/monitor-telegraf-service-account.yaml deleted file mode 100644 index ffa2622..0000000 --- a/charts/monitor/templates/monitor-telegraf-service-account.yaml +++ /dev/null @@ -1,6 +0,0 @@ -apiVersion: v1 -kind: ServiceAccount -metadata: - name: drycc-monitor-telegraf - labels: - heritage: drycc diff --git a/charts/monitor/values.yaml b/charts/monitor/values.yaml deleted file mode 100644 index 7d82c38..0000000 --- a/charts/monitor/values.yaml +++ /dev/null @@ -1,102 +0,0 @@ -grafana: - imageOrg: "drycc" - imagePullPolicy: "Always" - imageTag: "canary" - imageRegistry: "registry.drycc.cc" - # limitsCpu: "100m" - # limitsMemory: "50Mi" - nodeAffinityPreset: - key: "drycc.cc/node" - type: "soft" - values: - - "true" - - podAffinityPreset: - key: "security" - type: "" - values: - - "drycc-security" - - podAntiAffinityPreset: - key: "app" - type: "soft" - values: - - "drycc-monitor-grafana" - - # Service - service: - # Provide any additional service annotations - annotations: {} - - persistence: - enabled: false - accessMode: ReadWriteOnce - size: 5Gi - # Any custom grafana environment variables - # can be specified as key-value pairs under environment - # this is usually a non required setting. - environment: - ALLOW_SIGN_UP: false - -telegraf: - imageOrg: "drycc" - imagePullPolicy: "Always" - imageTag: "canary" - imageRegistry: "registry.drycc.cc" - # limitsCpu: "100m" - # limitsMemory: "50Mi" - nodeAffinityPreset: - key: "drycc.cc/node" - type: "soft" - values: - - "true" - - podAffinityPreset: - key: "security" - type: "" - values: - - "drycc-security" - - podAntiAffinityPreset: - key: "app" - type: "soft" - values: - - "drycc-monitor-telegraf" - -redis: - replicas: 1 - port: "6379" - -global: - # Set the location of Workflow's influxdb cluster - # - # Valid values are: - # - on-cluster: Run Influxdb within the Kubernetes cluster - # - off-cluster: Influxdb is running outside of the cluster and credentials and connection information will be provided. - influxdbLocation: "on-cluster" - # Set the location of Workflow's grafana instance - # - # Valid values are: - # - on-cluster: Run Grafana within the Kubernetes cluster - # - off-cluster: Grafana is running outside of the cluster - grafanaLocation: "on-cluster" - # Set the location of Workflow's redis instance - # - # Valid values are: - # - on-cluster: Run Redis within the Kubernetes cluster - # - off-cluster: Run Redis outside the Kubernetes cluster (configure in redis section) - redisLocation: "on-cluster" - # Role-Based Access Control for Kubernetes >= 1.5 - rbac: false - ingressClass: "" - # A domain name consists of one or more parts. - # Periods (.) are used to separate these parts. - # Each part must be 1 to 63 characters in length and can contain lowercase letters, digits, and hyphens (-). - # It must start and end with a lowercase letter or digit. - clusterDomain: "cluster.local" - # The public resolvable hostname to build your cluster with. - # - # This will be the hostname that is used to build endpoints such as "drycc.$HOSTNAME" - platformDomain: "" - # Whether cert_manager is enabled to automatically generate drycc certificates - certManagerEnabled: true diff --git a/default.mk b/default.mk deleted file mode 100644 index 1a2fde5..0000000 --- a/default.mk +++ /dev/null @@ -1,43 +0,0 @@ -SHELL := /bin/bash - -# all monitor components share/use the following targets/exports -DOCKER_HOST = $(shell echo $$DOCKER_HOST) -BUILD_TAG ?= git-$(shell git rev-parse --short HEAD) -DRYCC_REGISTRY ?= ${DEV_REGISTRY} -IMAGE_PREFIX ?= drycc -PLATFORM ?= linux/amd64,linux/arm64 - -include ../includes.mk -include ../versioning.mk -include ../deploy.mk - -TEST_ENV_PREFIX := docker run --rm -v ${CURDIR}:/bash -w /bash ${DEV_REGISTRY}/drycc/go-dev - -build: docker-build -push: docker-push -deploy: check-kubectl docker-build docker-push install - -docker-build: - docker build ${DOCKER_BUILD_FLAGS} -t ${IMAGE} rootfs - docker tag ${IMAGE} ${MUTABLE_IMAGE} - -docker-buildx: - docker buildx build --platform ${PLATFORM} -t ${IMAGE} rootfs --push - -clean: check-docker - docker rmi $(IMAGE) - -test: test-style - -test-style: - ${TEST_ENV_PREFIX} shellcheck $(SHELL_SCRIPTS) - -.PHONY: build push docker-build clean upgrade deploy test test-style - -build-all: - docker build ${DOCKER_BUILD_FLAGS} -t ${DRYCC_REGISTRY}/${IMAGE_PREFIX}/grafana:${VERSION} ../grafana/rootfs - docker build ${DOCKER_BUILD_FLAGS} -t ${DRYCC_REGISTRY}/${IMAGE_PREFIX}/telegraf:${VERSION} ../telegraf/rootfs - -push-all: - docker push ${DRYCC_REGISTRY}/${IMAGE_PREFIX}/grafana:${VERSION} - docker push ${DRYCC_REGISTRY}/${IMAGE_PREFIX}/telegraf:${VERSION} diff --git a/deploy.mk b/deploy.mk deleted file mode 100644 index 2864d5d..0000000 --- a/deploy.mk +++ /dev/null @@ -1,9 +0,0 @@ -install: - helm upgrade monitor ../charts/monitor --install --namespace drycc --set grafana.org=${IMAGE_PREFIX},grafana.image_tag=${VERSION},telegraf.org=${IMAGE_PREFIX},telegraf.image_tag=${VERSION} - - -upgrade: - helm upgrade monitor ../charts/monitor --namespace drycc --set grafana.org=${IMAGE_PREFIX},grafana.image_tag=${VERSION},telegraf.org=${IMAGE_PREFIX},telegraf.image_tag=${VERSION} - -uninstall: - helm delete monitor \ No newline at end of file diff --git a/grafana/Makefile b/grafana/Makefile deleted file mode 100644 index 8195e90..0000000 --- a/grafana/Makefile +++ /dev/null @@ -1,4 +0,0 @@ -SHORT_NAME ?= grafana -SHELL_SCRIPTS = rootfs/usr/share/grafana/start-grafana - -include ../default.mk diff --git a/grafana/README.md b/grafana/README.md deleted file mode 100644 index c690afc..0000000 --- a/grafana/README.md +++ /dev/null @@ -1,122 +0,0 @@ -# Grafana - -## Description -[Grafana](https://grafana.org/) is a graphing application built for time series data. It natively supports influxdb and provides great dashboarding support. This project is focused on provided a grafana installation that can be run within a kubernetes installation. The grafana application is agnostic to [Workflow](https://github.com/drycc/workflow) and can be installed as a stand alone system with the monitoring suite. - -## Configuration -| ENV Var | Default Value | Description | -|---------|---------------|-------------| -| [APP_MODE](#app_mode) | development | Has two possible values (production or development). | -| [DATA_PATH](#data_path) | /var/lib/grafana | Path to where grafana can store temp files, sessions, and the sqlite3 db (if that is used). | -| [LOG_PATH](#log_path) | /var/log | Directory where grafana can store logs. | -| [SERVER_PROTOCOL](#server_protocol) | http | Protocol (http or https). | -| [BIND_ADDRESS](#bind_address) | `""` | The ip address to bind to, empty will bind to all interfaces | -| [BIND_PORT](#bind_port) | 3000 | The http port to use | -| [DOMAIN](#domain) | localhost | The public facing domain name used to access grafana from a browser | -| [ENFORCE_DOMAIN](#enforce_domain) | false | Redirect to correct domain if host header does not match domain. Prevents DNS rebinding attacks | -| [ROOT_URL](#root_url) | `"%(protocol)s://%(domain)s:%(http_port)s/"` | The full public facing url | -| [ROUTER_LOGGING](#router_logging) | false | Log web requests | -| [STATIC_ROOT_PATH](#static_root_path) | public | the path relative working path | -| [ENABLE_GZIP](#enable_gzip) | false | enable gzip | -| [CERT_FILE](#cert_file) | no default | https certs | -| [CERT_KEY](#cert_key) | no default | https certs key | -| [DATABASE_TYPE](#database_type) | no default | Potential values: mysql, postgres, sqlite3 | -| [DATABASE_HOST](#database_host) | no default | Address to external database host | -| [DATABASE_NAME](#database_name) | no default | Name of database to store information | -| [DATABASE_USER](#database_user) | no default | User to log into the database with | -| [DATABASE_PASSWORD](#database_password) | no default | Database password | -| [DATABASE_SSL_MODE](#database_ssl_mode) | no default | For "postgres" only, either "disable", "require" or "verify-full" | -| [DATABASE_PATH](#database_path) | no default | For "sqlite3" only, path relative to data_path setting | -| [SESSION_PROVIDER](#session_provider) | file | Either "memory", "file", "redis", "mysql", "postgres", default is "file" | -| [SESSION_PROVIDER_CONFIG](#session_provider_config) | no default | Provider config options
`memory: not have any config yet`
`file: session dir path, is relative to grafana data_path`
`redis: addr=127.0.0.1:6379,pool_size=100,db=grafana`
`mysql: user:password@tcp(127.0.0.1:3306)/database_name`
`postgres: user=a password=b host=localhost port=5432 dbname=c sslmode=disable` | -| [SESSION_COOKIE_NAME](#session_cookie_name) | grafana_sess | Session cookie name | -| [SESSION_LIFE_TIME](#session_life_time) | 86400 | Session life time, default is 86400 | -| [GRAFANA_REPORTING](#grafana_reporting) | false | Server reporting, sends usage counters to stats.grafana.org every 24 hours. No ip addresses are being tracked, only simple counters to track running instances, dashboard and error counts. It is very helpful to us. Change this option to false to disable reporting. | -| [GOOGLE_ANALYTICS_UA_ID](#google_analytics_ua_id) | no default | Google Analytics universal tracking code, only enabled if you specify an id here | -| [DEFAULT_USER](#default_user) | admin | default admin user, created on startup | -| [DEFAULT_USER_PASSWORD](#default_user_password) | admin | default admin password, can be changed before first start of grafana, or in profile settings | -| [SECRET_KEY](#secret_key) | SW2YcwTIb9zpOOhoPsMm | used for signing | -| [LOGIN_REMEMBER_DAYS](#login_remember_day) | 7 | Auto-login remember days | -| [COOKIE_USERNAME](#cookie_username) | grafana_user | | -| [COOKIE_REMEMBER_NAME](#cookie_remember_name) | grafana_remember | | -| [DISABLE_GRAVATAR](#disable_gravatar) | false | disable gravatar profile images | -| [DATASOURCE_PROXY_WHITELIST](#datasource_proxy_whitelist) | no default | data source proxy whitelist (ip_or_domain:port seperated by spaces) | -| [ALLOW_SIGN_UP](#allow_sign_up) | true | allow user signup / registration | -| [ALLOW_ORG_CREATE](#allow_org_create) | true | Allow non admin users to create organizations | -| [AUTO_ASSIGN_ORG](#auto_assign_org) | true | Set to true to automatically assign new users to the default organization (id 1) | -| [AUTO_ASSIGN_ORG_ROLE](#auto_assign_org_role) | Viewer | Default role new users will be automatically assigned (if disabled above is set to true) | -| [LOGIN_HINT](#login_hint) | email or username | Background text for the user field on the login page | -| [ANON_ACCESS](#anon_access) | false | enable anonymous access | -| [ORG_NAME](#org_name) | Main Org. | specify organization name that should be used for unauthenticated users | -| [UNAUTHED_USER_ORG_ROLE](#unauthed_user_org_role) | Viewer | specify role for unauthenticated users | -| [GITHUB_AUTH](#github_auth) | no default | Enable Github Auth (true/false) | -| [GITHUB_AUTH_ALLOW_SIGN_UP](#github_auth_allow_sign_up) | false | Allow signup with github auth | -| [GITHUB_AUTH_CLIENT_ID](#github_auth_client_id) | no default | Github Client Auth | -| [GITHUB_AUTH_CLIENT_SECRET](#github_auth_client_secret) | no default | Github Auth client secret | -| [GITHUB_AUTH_SCOPES](#github_auth_scopes) | user:email,read:org | Github Auth Scopes | -| [GITHUB_AUTH_URL](#github_auth_url) | https://github.com/login/oauth/authorize | Github authorization URL | -| [GITHUB_AUTH_TOKEN_URL](#github_auth_token_url) | https://github.com/login/oauth/access_token | Github Auth Token URL | -| [GITHUB_AUTH_API_URL](#github_auth_api_url) | https://api.github.com/user | Github Auth API URL | -| [GITHUB_AUTH_TEAM_IDS](#github_auth_team_ids) | "" | Team IDs to use for Github Auth | -| [GITHUB_AUTH_ALLOWED_ORGS](#github_auth_allowed_orgs) | "" | Allowed ORGs for Github Auth | -| [GOOGLE_AUTH](#google_auth) | no default | Enable Google Auth for login (true/false) | -| [GOOGLE_AUTH_ALLOW_SIGN_UP](#google_auth_allow_sign_up) | false | Allow people to sign up using Google Auth | -| [GOOGLE_AUTH_CLIENT_ID](#google_auth_client_id) | no default | Google Auth Client ID | -| [GOOGLE_AUTH_CLIENT_SECRET](#google_auth_client_secret) | no default | Google Auth Client Secret | -| [GOOGLE_AUTH_SCOPES](#google_auth_scopes) | https://www.googleapis.com/auth/userinfo.profile
https://www.googleapis.com/auth/userinfo.email | Google Auth Scopes | -| [GOOGLE_AUTH_URL](#google_auth_url) | https://accounts.google.com/o/oauth2/auth | Google Auth URL | -| [GOOGLE_AUTH_TOKEN_URL](#google_auth_token_url) | https://accounts.google.com/o/oauth2/token | Google Auth Token URL | -| [GOOGLE_AUTH_API_URL](#google_auth_api_url) | https://www.googleapis.com/oauth2/v1/userinfo | Google Auth API URL | -| [GOOGLE_AUTH_ALLOWED_DOMAINS](#google_auth_allowed_domains) | "" | Google Auth allowed domains | -| [AUTH_PROXY](#auth_proxy) | no default | If enabled use an authorization proxy | -| [AUTH_HEADER_NAME](#auth_header_name) | X-WEBAUTH-USER | Header to use for authorization | -| [AUTH_HEADER_PROPERTY](#auth_header_property) | username | Property to use in Authorization Header | -| [AUTH_AUTO_SIGN_UP](#auth_auto_sign_up) | true | Auto Signup user with data passed from Authorization Header| -| [BASIC_AUTH](#basic_auth) | true | Enable Basic Auth | -| [LDAP_AUTH](#ldap_auth) | no default | Enable LDAP Auth | -| [LDAP_AUTH_CONFIG_FILE](#ldap_auth_config_file) | /etc/grafana/ldap.toml | Config file to use for LDAP Auth | -| [SMTP](#smtp) | no default | Enable SMTP/Emailing | -| [SMTP_HOST](#smtp_host) | no default | SMTP Host | -| [SMTP_USER](#smtp_user) | no default | User for SMTP | -| [SMTP_PASSWORD](#smtp_password) | no default | Password for SMTP | -| [SMTP_CERT_FILE](#smtp_cert_file) | no default | Cert file for SMTP HTTPS | -| [SMTP_KEY_FILE](#smtp_key_file) | no default | Key file for SMTP HTTPS | -| [SMTP_SKIP_VERIFY](#smtp_skip_verify) | false | Skip HTTPS verify | -| [SMTP_FROM_ADDRESS](#smtp_from_address) | admin@grafana.localhost | From address to use when sending emails | -| [WELCOME_EMAIL](#welcome_email) | false | Send welcome email on sign up | -| [LOG_MODE](#log_mode) | console | Either "console", "file", default is "console" Use comma to separate multiple modes, e.g. "console, file" | -| [BUFFER_LENGTH](#buffer_length) | 10000 | Buffer length of channel, keep it as it is if you don't know what it is. | -| [LOG_LEVEL](#log_level) | Info | Either "Trace", "Debug", "Info", "Warn", "Error", "Critical", default is "Trace" | -| [LOG_LEVEL_CONSOLE](#log_level_console) | Info | For "console" mode only | -| [LOG_LEVEL_FILE](#log_level_file) | Info | For "file" mode only | -| [LOG_ROTATE](#log_rotate) | true | This enables automated log rotate(switch of following options), default is true | -| [LOG_MAX_LINES](#log_max_lines) | 1000000 | Max line number of single file | -| [LOG_MAX_LINES_SHIFT](#log_max_lines_shift) | 28 | Max size shift of single file, default is 28 means 1 << 28, 256MB | -| [LOG_DAILY_ROTATE](#log_daily_rotate) | true | Segment log daily, default is true | -| [LOG_MAX_DAYS](#log_max_days) | 7 | Expired days of log file(delete after max days) | -| [AMQP_PUBLISHER](#amqp_publisher) | no default | Publish Events to AMQP | -| [AMQP_PUBLISHER_HOST](#amqp_publisher_host) | no default | RabbitMQ URL | -| [AMQP_PUBLISHER_EXCHANGE](#amqp_publisher_exchange) | no default | AMQP Publisher Exchange | -| [DASHBOARD_JSON](#dashboard_json) | true | Poll a location for json files that contain dashboards | -| [DASHBOARD_JSON_PATH](#dashboard_json_path) | /usr/share/grafana/dashboards | Location to scan for json dashboards | -| [PLUGINS_PATH](#plugins_path) | /var/lib/grafana/plugins | Path to where grafana can install plugins | -| [GF_INSTALL_PLUGINS](#gf_install_plugins) | no default | Pass the plugins as a comma seperated list | - -## Development -The provided `Makefile` has various targets to help support building and publishing new images into a kubernetes cluster. - -### Environment variables -There are a few key environment variables you should be aware of when interacting with the `make` targets. - -* `BUILD_TAG` - The tag provided to the docker image when it is built (defaults to the git-sha) -* `SHORT_NAME` - The name of the image (defaults to `grafana`) -* `DRYCC_REGISTRY` - This is the registry you are using (default `dockerhub`) -* `IMAGE_PREFIX` - This is the account for the registry you are using (default `drycc`) - -### Make targets - -* `make build` - Build docker image -* `make push` - Push docker image to a registry -* `make upgrade` - Replaces the running grafana instance with a new one - -The typical workflow will look something like this - `DRYCC_REGISTRY= IMAGE_PREFIX=foouser make build push upgrade` diff --git a/grafana/rootfs/Dockerfile b/grafana/rootfs/Dockerfile deleted file mode 100644 index 951d8ed..0000000 --- a/grafana/rootfs/Dockerfile +++ /dev/null @@ -1,40 +0,0 @@ -FROM registry.drycc.cc/drycc/base:bullseye - -ENV DRYCC_UID=1001 \ - DRYCC_GID=1001 \ - DRYCC_HOME_DIR=/usr/share/grafana \ - GRAFANA_VERSION="9.1.1" \ - ENVTPL_VERSION="1.0.0" - -RUN groupadd drycc --gid ${DRYCC_GID} \ - && useradd drycc -u ${DRYCC_UID} -g ${DRYCC_GID} -s /bin/bash -m -d ${DRYCC_HOME_DIR} - -COPY . / - -RUN install-packages fontconfig curl \ - && install-stack grafana ${GRAFANA_VERSION} \ - && install-stack envtpl ${ENVTPL_VERSION} \ - && rm -rf \ - /usr/share/doc \ - /usr/share/man \ - /usr/share/info \ - /usr/share/locale \ - /var/lib/apt/lists/* \ - /var/log/* \ - /var/cache/debconf/* \ - /etc/systemd \ - /lib/lsb \ - /lib/udev \ - /usr/lib/`echo $(uname -m)`-linux-gnu/gconv/IBM* \ - /usr/lib/`echo $(uname -m)`-linux-gnu/gconv/EBC* \ - && mkdir -p /usr/share/man/man{1..8} \ - && mkdir /var/lib/grafana -p \ - && chown -R ${DRYCC_UID}:${DRYCC_GID} /opt/drycc/grafana /usr/share/grafana /var/lib/grafana \ - && chmod +x /usr/share/grafana/start-grafana \ - && chmod 644 /usr/share/grafana/grafana.ini.tpl \ - && chmod +x /usr/share/grafana/entrypoint.sh - -USER ${DRYCC_UID} -WORKDIR ${DRYCC_HOME_DIR} -CMD ["./start-grafana"] -EXPOSE 3000 diff --git a/grafana/rootfs/usr/share/grafana/api/dashboards/drycc_component_health.json b/grafana/rootfs/usr/share/grafana/api/dashboards/drycc_component_health.json deleted file mode 100644 index cd65592..0000000 --- a/grafana/rootfs/usr/share/grafana/api/dashboards/drycc_component_health.json +++ /dev/null @@ -1,4241 +0,0 @@ -{ - "dashboard": { - "annotations": { - "list": [ - { - "builtIn": 1, - "datasource": "-- Grafana --", - "enable": true, - "hide": true, - "iconColor": "rgba(0, 211, 255, 1)", - "name": "Annotations & Alerts", - "target": { - "limit": 100, - "matchAny": false, - "tags": [], - "type": "dashboard" - }, - "type": "dashboard" - } - ] - }, - "editable": true, - "fiscalYearStartMonth": 0, - "graphTooltip": 0, - "id": null, - "links": [], - "liveNow": false, - "panels": [ - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 0 - }, - "id": 33, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 3 - }, - "hiddenSeries": false, - "id": 31, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_namespace", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "namespace" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> keep(columns: [\"_time\", \"namespace\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: sum, createEmpty: false)\n |> yield(name: \"sum\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Total CPU Usage", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 3 - }, - "hiddenSeries": false, - "id": 32, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_namespace", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "namespace" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> keep(columns: [\"_time\", \"namespace\", \"_value\"])\n |> movingAverage(n: 10)\n |> aggregateWindow(every: v.windowPeriod, fn: sum, createEmpty: false)\n |> yield(name: \"sum\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Total Memory Usage", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "SUM", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 1 - }, - "id": 52, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 4 - }, - "hiddenSeries": false, - "id": 54, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "hide": false, - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-passport\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 4 - }, - "hiddenSeries": false, - "id": 56, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-passport\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "PASSPORT", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 2 - }, - "id": 34, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 4 - }, - "hiddenSeries": false, - "id": 1, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "hide": false, - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] =~ /drycc-controller/)\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 4 - }, - "hiddenSeries": false, - "id": 2, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] =~ /drycc-controller/)\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "CONTROLLER", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 3 - }, - "id": 58, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 3 - }, - "hiddenSeries": false, - "id": 60, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "hide": false, - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] =~ /drycc-helmbroker/)\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 3 - }, - "hiddenSeries": false, - "id": 62, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] =~ /drycc-helmbroker/)\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "HELM-BROKER", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 4 - }, - "id": 35, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 5 - }, - "hiddenSeries": false, - "id": 3, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-builder\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 5 - }, - "hiddenSeries": false, - "id": 4, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-builder\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "BUILDER", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 5 - }, - "id": 36, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 6 - }, - "hiddenSeries": false, - "id": 6, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-database\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 6 - }, - "hiddenSeries": false, - "id": 5, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-database\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "DATABASE", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 6 - }, - "id": 37, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 7 - }, - "hiddenSeries": false, - "id": 19, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-logger-fluentd\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 7 - }, - "hiddenSeries": false, - "id": 20, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-logger-fluentd\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "FLUENTD", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 7 - }, - "id": 38, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 8 - }, - "hiddenSeries": false, - "id": 25, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-monitor-grafana\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 8 - }, - "hiddenSeries": false, - "id": 26, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-monitor-grafana\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "GRAFANA", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 8 - }, - "id": 39, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 9 - }, - "hiddenSeries": false, - "id": 23, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] =~ /drycc-influxdb/)\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 9 - }, - "hiddenSeries": false, - "id": 24, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] =~ /drycc-influxdb/)\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "INFLUXDB", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 9 - }, - "id": 40, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 10 - }, - "hiddenSeries": false, - "id": 7, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-logger\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 10 - }, - "hiddenSeries": false, - "id": 8, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-logger\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "LOGGER", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 10 - }, - "id": 41, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 11 - }, - "hiddenSeries": false, - "id": 9, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-storage-minio\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 11 - }, - "hiddenSeries": false, - "id": 10, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-storage-minio\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "STORAGE-MINIO", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 11 - }, - "id": 64, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 2 - }, - "hiddenSeries": false, - "id": 68, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "hide": false, - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-storage-meta-pd\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 2 - }, - "hiddenSeries": false, - "id": 70, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-storage-meta-pd\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "STORAGE-META-PD", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 12 - }, - "id": 66, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 1 - }, - "hiddenSeries": false, - "id": 72, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "hide": false, - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-storage-meta-tikv\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 1 - }, - "hiddenSeries": false, - "id": 74, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-storage-meta-tikv\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "STORAGE-META-TIKV", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 13 - }, - "id": 42, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 12 - }, - "hiddenSeries": false, - "id": 13, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-rabbitmq\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 12 - }, - "hiddenSeries": false, - "id": 14, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-rabbitmq\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "RABBITMQ", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 14 - }, - "id": 43, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 13 - }, - "hiddenSeries": false, - "id": 27, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-redis\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 13 - }, - "hiddenSeries": false, - "id": 28, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-redis\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "REDIS", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 15 - }, - "id": 44, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 14 - }, - "hiddenSeries": false, - "id": 11, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-registry\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 14 - }, - "hiddenSeries": false, - "id": 12, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-registry\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "REGISTRY", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 16 - }, - "id": 45, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 15 - }, - "hiddenSeries": false, - "id": 16, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-registry-proxy\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 15 - }, - "hiddenSeries": false, - "id": 15, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-registry-proxy\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "REGISTRY-PROXY", - "type": "row" - }, - { - "collapsed": true, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 17 - }, - "id": 46, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 2, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 16 - }, - "hiddenSeries": false, - "id": 21, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-monitor-telegraf\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 1, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 16 - }, - "hiddenSeries": false, - "id": 22, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-monitor-telegraf\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "title": "TELEGRAF", - "type": "row" - } - ], - "refresh": "5s", - "schemaVersion": 35, - "style": "dark", - "tags": [ - "drycc" - ], - "templating": { - "list": [] - }, - "time": { - "from": "now-3h", - "to": "now" - }, - "timepicker": { - "refresh_intervals": [ - "5s", - "10s", - "30s", - "1m", - "5m", - "15m", - "30m", - "1h", - "2h", - "1d" - ], - "time_options": [ - "5m", - "15m", - "1h", - "6h", - "12h", - "24h", - "2d", - "7d", - "30d" - ] - }, - "timezone": "browser", - "title": "Drycc Component Health", - "uid": "2ySBrTTMz", - "version": 10, - "weekStart": "" - }, - "overwrite": false -} \ No newline at end of file diff --git a/grafana/rootfs/usr/share/grafana/api/dashboards/influx.json b/grafana/rootfs/usr/share/grafana/api/dashboards/influx.json deleted file mode 100644 index a257b8b..0000000 --- a/grafana/rootfs/usr/share/grafana/api/dashboards/influx.json +++ /dev/null @@ -1,619 +0,0 @@ -{ - "dashboard": { - "annotations": { - "list": [ - { - "builtIn": 1, - "datasource": { - "type": "datasource", - "uid": "grafana" - }, - "enable": true, - "hide": true, - "iconColor": "rgba(0, 211, 255, 1)", - "name": "Annotations & Alerts", - "target": { - "limit": 100, - "matchAny": false, - "tags": [], - "type": "dashboard" - }, - "type": "dashboard" - } - ] - }, - "editable": true, - "fiscalYearStartMonth": 0, - "graphTooltip": 0, - "id": null, - "links": [], - "liveNow": false, - "panels": [ - { - "datasource": null, - "fieldConfig": { - "defaults": { - "color": { - "mode": "thresholds" - }, - "mappings": [ - { - "options": { - "match": "null", - "result": { - "text": "N/A" - } - }, - "type": "special" - } - ], - "thresholds": { - "mode": "absolute", - "steps": [ - { - "color": "green", - "value": null - }, - { - "color": "red", - "value": 80 - } - ] - }, - "unit": "none" - }, - "overrides": [] - }, - "gridPos": { - "h": 3, - "w": 12, - "x": 0, - "y": 0 - }, - "id": 2, - "links": [], - "maxDataPoints": 100, - "options": { - "colorMode": "none", - "graphMode": "none", - "justifyMode": "auto", - "orientation": "horizontal", - "reduceOptions": { - "calcs": [ - "mean" - ], - "fields": "", - "values": false - }, - "textMode": "auto" - }, - "pluginVersion": "9.1.1", - "targets": [ - { - "datasource": null, - "groupBy": [ - { - "params": [ - "$__interval" - ], - "type": "time" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "orderByTime": "ASC", - "policy": "default", - "query": "import \"influxdata/influxdb/schema\"\nschema.measurements(bucket: v.bucket)\n|> count()", - "refId": "A", - "resultFormat": "time_series", - "select": [ - [ - { - "params": [ - "value" - ], - "type": "field" - }, - { - "params": [], - "type": "mean" - } - ] - ], - "tags": [] - } - ], - "title": "Number of Measurements", - "type": "stat" - }, - { - "datasource": null, - "fieldConfig": { - "defaults": { - "color": { - "fixedColor": "rgb(31, 120, 193)", - "mode": "fixed" - }, - "mappings": [ - { - "options": { - "match": "null", - "result": { - "text": "N/A" - } - }, - "type": "special" - } - ], - "thresholds": { - "mode": "absolute", - "steps": [ - { - "color": "green", - "value": null - }, - { - "color": "red", - "value": 80 - } - ] - }, - "unit": "none" - }, - "overrides": [] - }, - "gridPos": { - "h": 3, - "w": 12, - "x": 12, - "y": 0 - }, - "id": 5, - "links": [], - "maxDataPoints": 100, - "options": { - "colorMode": "none", - "graphMode": "area", - "justifyMode": "auto", - "orientation": "horizontal", - "reduceOptions": { - "calcs": [ - "lastNotNull" - ], - "fields": "", - "values": false - }, - "textMode": "auto" - }, - "pluginVersion": "9.1.1", - "targets": [ - { - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "influxdb_shard", - "orderByTime": "ASC", - "policy": "default", - "query": "import \"csv\"\n\ndummy_data = \"#group,false,false,false,false,true,true,true,true\n#datatype,string,long,dateTime:RFC3339,double,string,string,string,string\n#default,_result,,,,,,,\n,result,table,_time,_value,_field,_measurement,status,task_type\n,,0,2000-01-01T00:00:00.000000Z,0,counter,task_executor_total_runs_complete,success,system\"\n\ndefault = csv.from(csv: dummy_data)\n\nnew = from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r._measurement == \"task_executor_total_runs_complete\")\n |> filter(fn: (r) => r.status == \"success\")\n |> filter(fn: (r) => r.task_type == \"system\")\n |> filter(fn: (r) => r._field == \"counter\")\n |> last()\n\nunion(tables: [default, new]) |> last()", - "refId": "A", - "resultFormat": "time_series", - "select": [ - [ - { - "params": [ - "diskBytes" - ], - "type": "field" - }, - { - "params": [], - "type": "last" - } - ] - ], - "tags": [] - } - ], - "title": "Successful Task Runs", - "type": "stat" - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": null, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 6, - "w": 12, - "x": 0, - "y": 3 - }, - "hiddenSeries": false, - "id": 7, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "9.1.1", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "hide": false, - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"storage_shard_disk_size\")\n |> keep(columns: [\"_time\", \"_value\", \"host\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Share Disk Size", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": null, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 6, - "w": 12, - "x": 12, - "y": 3 - }, - "hiddenSeries": false, - "id": 9, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "9.1.1", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "hide": false, - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"influxdb_buckets_total\")\n |> filter(fn: (r) => r[\"_field\"] == \"counter\")\n |> keep(columns: [\"_time\", \"_value\", \"host\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Number of Buckets", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": null, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 24, - "x": 0, - "y": 9 - }, - "hiddenSeries": false, - "id": 11, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "9.1.1", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "hide": false, - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r._measurement == \"boltdb_reads_total\")\n |> derivative(unit: 1s, nonNegative: true)\n |> drop(columns: [\"url\", \"_field\", \"_measurement\"])", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Local Object Store IO", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "Millicores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "schemaVersion": 37, - "style": "dark", - "tags": [ - "drycc" - ], - "templating": { - "list": [] - }, - "time": { - "from": "now-3h", - "to": "now" - }, - "timepicker": { - "now": true, - "refresh_intervals": [ - "5s", - "10s", - "30s", - "1m", - "5m", - "15m", - "30m", - "1h", - "2h", - "1d" - ], - "time_options": [ - "5m", - "15m", - "1h", - "6h", - "12h", - "24h", - "2d", - "7d", - "30d" - ] - }, - "timezone": "browser", - "title": "Influx", - "uid": "nUSfrTTMk", - "version": 0, - "weekStart": "" - }, - "overwrite": false -} \ No newline at end of file diff --git a/grafana/rootfs/usr/share/grafana/api/dashboards/kube_health.json b/grafana/rootfs/usr/share/grafana/api/dashboards/kube_health.json deleted file mode 100644 index 0cf95d4..0000000 --- a/grafana/rootfs/usr/share/grafana/api/dashboards/kube_health.json +++ /dev/null @@ -1,962 +0,0 @@ -{ - "dashboard": { - "annotations": { - "list": [ - { - "builtIn": 1, - "datasource": "-- Grafana --", - "enable": true, - "hide": true, - "iconColor": "rgba(0, 211, 255, 1)", - "name": "Annotations & Alerts", - "target": { - "limit": 100, - "matchAny": false, - "tags": [], - "type": "dashboard" - }, - "type": "dashboard" - } - ] - }, - "editable": true, - "fiscalYearStartMonth": 0, - "graphTooltip": 0, - "id": null, - "iteration": 1660728936662, - "links": [], - "liveNow": false, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 24, - "x": 0, - "y": 0 - }, - "hiddenSeries": false, - "id": 1, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "import \"math\"\nfrom(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"kubernetes_pod_container\")\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"container_name\"] =~ /^$container_name$/)\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "milli-cores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 24, - "x": 0, - "y": 7 - }, - "hiddenSeries": false, - "id": 2, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "import \"math\"\nfrom(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"kubernetes_pod_container\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> filter(fn: (r) => r[\"container_name\"] =~ /^$container_name$/)\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value\n })\n )\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "editable": true, - "error": false, - "gridPos": { - "h": 7, - "w": 2, - "x": 0, - "y": 14 - }, - "id": 5, - "isNew": true, - "links": [], - "options": { - "content": "# Kubelet", - "mode": "markdown" - }, - "pluginVersion": "8.4.5", - "transparent": true, - "type": "text" - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 22, - "x": 2, - "y": 14 - }, - "hiddenSeries": false, - "id": 4, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_system_container", - "orderByTime": "ASC", - "policy": "default", - "query": "import \"math\"\nfrom(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"kubernetes_system_container\")\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"container_name\"] == \"kubelet\")\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "milli-cores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "editable": true, - "error": false, - "gridPos": { - "h": 7, - "w": 2, - "x": 0, - "y": 21 - }, - "id": 8, - "isNew": true, - "links": [], - "options": { - "content": "", - "mode": "markdown" - }, - "pluginVersion": "8.4.5", - "transparent": true, - "type": "text" - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 22, - "x": 2, - "y": 21 - }, - "hiddenSeries": false, - "id": 6, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_system_container", - "orderByTime": "ASC", - "policy": "default", - "query": "import \"math\"\nfrom(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"kubernetes_system_container\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> filter(fn: (r) => r[\"container_name\"] == \"kubelet\" )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> movingAverage(n: 10)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "label": "milli-cores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "editable": true, - "error": false, - "gridPos": { - "h": 7, - "w": 2, - "x": 0, - "y": 28 - }, - "id": 7, - "isNew": true, - "links": [], - "options": { - "content": "# Kube System", - "mode": "markdown" - }, - "pluginVersion": "8.4.5", - "transparent": true, - "type": "text" - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 22, - "x": 2, - "y": 28 - }, - "hiddenSeries": false, - "id": 3, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - { - "alias": "Received", - "color": "#3F6833" - }, - { - "alias": "Sent", - "color": "#890F02", - "transform": "negative-Y" - } - ], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "Received", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_network", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"kubernetes_pod_network\" and r[\"_field\"] == \"rx_bytes\" and r[\"namespace\"] == \"kube-system\")\n |> derivative(unit: 10s, nonNegative: true)\n |> group(columns: [\"_time\"])\n |> sum()\n |> set(key: \"_field\", value: \"Received\")\n |> group(columns: [\"_value\", \"_time\"], mode: \"except\")\n |> aggregateWindow(every: v.windowPeriod, fn: sum, createEmpty: false)\n |> yield(name: \"mean\")\n", - "refId": "A", - "resultFormat": "time_series" - }, - { - "groupBy": [ - { - "params": [ - "$__interval" - ], - "type": "time" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "hide": false, - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"kubernetes_pod_network\" and r[\"_field\"] == \"tx_bytes\" and r[\"namespace\"] == \"kube-system\")\n |> derivative(unit: 10s, nonNegative: true)\n |> group(columns: [\"_time\"])\n |> sum()\n |> set(key: \"_field\", value: \"Sent\")\n |> group(columns: [\"_value\", \"_time\"], mode: \"except\")\n |> aggregateWindow(every: v.windowPeriod, fn: sum, createEmpty: false)\n |> yield(name: \"mean\")\n", - "refId": "B", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Network", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "Bps", - "label": "", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "editable": true, - "error": false, - "gridPos": { - "h": 7, - "w": 2, - "x": 0, - "y": 35 - }, - "id": 10, - "isNew": true, - "links": [], - "options": { - "content": "# Nodes", - "mode": "markdown" - }, - "pluginVersion": "8.4.5", - "transparent": true, - "type": "text" - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 22, - "x": 2, - "y": 35 - }, - "hiddenSeries": false, - "id": 9, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_node_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "node_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_node", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"kubernetes_node\")\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> group(columns: [\"host\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "label": "milli-cores", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "editable": true, - "error": false, - "gridPos": { - "h": 7, - "w": 2, - "x": 0, - "y": 42 - }, - "id": 12, - "isNew": true, - "links": [], - "options": { - "content": "", - "mode": "markdown" - }, - "pluginVersion": "8.4.5", - "transparent": true, - "type": "text" - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 22, - "x": 2, - "y": 42 - }, - "hiddenSeries": false, - "id": 11, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "8.4.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_node_name", - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "node_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_node", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"kubernetes_node\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> group(columns: [\"host\"])\n |> movingAverage(n: 10)\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "label": "", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "refresh": "5s", - "schemaVersion": 35, - "style": "dark", - "tags": [], - "templating": { - "list": [ - { - "current": { - "selected": true, - "text": "cilium-agent", - "value": "cilium-agent" - }, - "definition": "import \"influxdata/influxdb/schema\" schema.tagValues(bucket: v.bucket, tag: \"container_name\", predicate: (r) => r.namespace =~ /kube-*/ or r.container_name =~ /kube.*/ )", - "hide": 0, - "includeAll": false, - "multi": false, - "name": "container_name", - "options": [], - "query": "import \"influxdata/influxdb/schema\" schema.tagValues(bucket: v.bucket, tag: \"container_name\", predicate: (r) => r.namespace =~ /kube-*/ or r.container_name =~ /kube.*/ )", - "refresh": 1, - "regex": "", - "skipUrlSync": false, - "sort": 0, - "type": "query" - } - ] - }, - "time": { - "from": "now-3h", - "to": "now" - }, - "timepicker": { - "refresh_intervals": [ - "5s", - "10s", - "30s", - "1m", - "5m", - "15m", - "30m", - "1h", - "2h", - "1d" - ], - "time_options": [ - "5m", - "15m", - "1h", - "6h", - "12h", - "24h", - "2d", - "7d", - "30d" - ] - }, - "timezone": "browser", - "title": "Kubernetes Health", - "uid": "X8IfrooGk", - "version": 2, - "weekStart": "" - }, - "overwrite": false -} \ No newline at end of file diff --git a/grafana/rootfs/usr/share/grafana/api/dashboards/redis.json b/grafana/rootfs/usr/share/grafana/api/dashboards/redis.json deleted file mode 100644 index 35b8eff..0000000 --- a/grafana/rootfs/usr/share/grafana/api/dashboards/redis.json +++ /dev/null @@ -1,507 +0,0 @@ -{ - "dashboard": { - "annotations": { - "list": [ - { - "builtIn": 1, - "datasource": "-- Grafana --", - "enable": true, - "hide": true, - "iconColor": "rgba(0, 211, 255, 1)", - "name": "Annotations & Alerts", - "target": { - "limit": 100, - "matchAny": false, - "tags": [], - "type": "dashboard" - }, - "type": "dashboard" - } - ] - }, - "editable": true, - "fiscalYearStartMonth": 0, - "graphTooltip": 0, - "id": null, - "links": [], - "liveNow": false, - "panels": [ - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": null, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 0 - }, - "hiddenSeries": false, - "id": 8, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "9.1.1", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_field\"] == \"cpu_usage_nanocores\")\n |> filter(fn: (r) => r[\"namespace\"] == \"drycc\")\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-redis\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> movingAverage(n: 10)\n |> map(fn: (r) => ({\n r with\n _value: r._value / 1000000.0\n })\n )\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "CPU Usage (Millicores)", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": null, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 0 - }, - "hiddenSeries": false, - "id": 10, - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "9.1.1", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"container_name\"] == \"drycc-redis\")\n |> filter(fn: (r) => r[\"_field\"] == \"memory_usage_bytes\")\n |> keep(columns: [\"_time\", \"node_name\", \"pod_name\", \"container_name\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Memory", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "bytes", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": null, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 0, - "y": 7 - }, - "hiddenSeries": false, - "id": 12, - "interval": "", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "9.1.1", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "alias": "$tag_pod_name", - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "pod_name" - ], - "type": "tag" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "kubernetes_pod_container", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"redis\")\n |> filter(fn: (r) => r[\"_field\"] == \"clients\")\n |> keep(columns: [\"_time\", \"host\", \"server\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Connected Clients", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "short", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": null, - "editable": true, - "error": false, - "fill": 1, - "fillGradient": 0, - "grid": {}, - "gridPos": { - "h": 7, - "w": 12, - "x": 12, - "y": 7 - }, - "hiddenSeries": false, - "id": 4, - "interval": "1s", - "isNew": true, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": false, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "connected", - "options": { - "alertThreshold": true - }, - "percentage": false, - "pluginVersion": "9.1.1", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "datasource": null, - "dsType": "influxdb", - "groupBy": [ - { - "params": [ - "$interval" - ], - "type": "time" - }, - { - "params": [ - "null" - ], - "type": "fill" - } - ], - "measurement": "redis", - "orderByTime": "ASC", - "policy": "default", - "query": "from(bucket: v.bucket)\n |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n |> filter(fn: (r) => r[\"_measurement\"] == \"redis\")\n |> filter(fn: (r) => r[\"_field\"] == \"instantaneous_ops_per_sec\")\n |> keep(columns: [\"_time\", \"host\", \"server\", \"_value\"])\n |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)\n |> yield(name: \"mean\")\n", - "refId": "A", - "resultFormat": "time_series" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Ops Per Second", - "tooltip": { - "msResolution": true, - "shared": true, - "sort": 0, - "value_type": "cumulative" - }, - "type": "graph", - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "yaxes": [ - { - "format": "ops", - "logBase": 1, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": true - } - ], - "yaxis": { - "align": false - } - } - ], - "refresh": "5s", - "schemaVersion": 37, - "style": "dark", - "tags": [ - "drycc" - ], - "templating": { - "list": [] - }, - "time": { - "from": "now-3h", - "to": "now" - }, - "timepicker": { - "refresh_intervals": [ - "5s", - "10s", - "30s", - "1m", - "5m", - "15m", - "30m", - "1h", - "2h", - "1d" - ], - "time_options": [ - "5m", - "15m", - "1h", - "6h", - "12h", - "24h", - "2d", - "7d", - "30d" - ] - }, - "timezone": "browser", - "title": "Redis", - "uid": "FlIfrooGz", - "version": 0, - "weekStart": "" - }, - "overwrite": false -} \ No newline at end of file diff --git a/grafana/rootfs/usr/share/grafana/entrypoint.sh b/grafana/rootfs/usr/share/grafana/entrypoint.sh deleted file mode 100644 index 5fc4448..0000000 --- a/grafana/rootfs/usr/share/grafana/entrypoint.sh +++ /dev/null @@ -1,3 +0,0 @@ -#!/bin/bash - -exec "$@" diff --git a/grafana/rootfs/usr/share/grafana/grafana.ini.tpl b/grafana/rootfs/usr/share/grafana/grafana.ini.tpl deleted file mode 100644 index ffbb2e8..0000000 --- a/grafana/rootfs/usr/share/grafana/grafana.ini.tpl +++ /dev/null @@ -1,314 +0,0 @@ -#################### Grafana Configuration Example ##################### -# -# Everything has defaults so you only need to uncomment things you want to -# change - -# possible values : production, development -app_mode = {{ default "development" .APP_MODE }} - -#################################### Paths #################################### -[paths] -# Path to where grafana can store temp files, sessions, and the sqlite3 db (if that is used) -# -data = {{ default "/var/lib/grafana" .DATA_PATH }} - -# -# Directory where grafana can store logs -# -logs = {{ default "/var/log" .LOG_PATH }} - -# -# Directory where grafana will automatically scan and look for plugins -# -plugins = {{ default "/var/lib/grafana/plugins" .PLUGINS_PATH }} - -#################################### Server #################################### -[server] -# Protocol (http or https) -protocol = {{ default "http" .SERVER_PROTOCOL }} - -# The ip address to bind to, empty will bind to all interfaces -http_addr = {{ default "" .BIND_ADDRESS}} -# The http port to use -http_port = {{ default "3000" .BIND_PORT }} - -# The public facing domain name used to access grafana from a browser -domain = {{ default "localhost" .DOMAIN }} - -# Redirect to correct domain if host header does not match domain -# Prevents DNS rebinding attacks -enforce_domain = {{ default "false" .ENFORCE_DOMAIN }} - -# The full public facing url -root_url = {{ default "%(protocol)s://%(domain)s:%(http_port)s/" .ROOT_URL }} - -# Log web requests -router_logging = {{ default "false" .ROUTER_LOGGING }} - -# the path relative working path -static_root_path = {{ default "public" .STATIC_ROOT_PATH }} - -# enable gzip -enable_gzip = {{ default "false" .ENABLE_GZIP }} - -# https certs & key file -{{ if .CERT_FILE }} -cert_file = {{ .CERT_FILE }} -{{ end }} -{{ if .CERT_KEY }} -cert_key = {{ .CERT_KEY }} -{{ end }} - -#################################### Database #################################### -[database] -# Either "mysql", "postgres" or "sqlite3", it's your choice -{{ if .DATABASE_TYPE }} -type = {{ .DATABASE_TYPE }} -{{ end }} -{{ if .DATABASE_HOST }} -host = {{ .DATABASE_HOST }} -{{ end }} -{{ if .DATABASE_NAME }} -name = {{ .DATABASE_NAME }} -{{ end }} -{{ if .DATABASE_USER }} -user = {{ .DATABASE_USER }} -{{ end }} -{{ if .DATABASE_PASSWORD }} -password = {{ .DATABASE_PASSWORD }} -{{ end }} - -# For "postgres" only, either "disable", "require" or "verify-full" -{{ if .DATABASE_SSL_MODE }} -ssl_mode = {{ .DATABASE_SSL_MODE }} -{{ end }} - -# For "sqlite3" only, path relative to data_path setting -{{ if .DATABASE_PATH }} -path = {{ .DATABASE_PATH }} -{{ end }} - -#################################### Session #################################### -[session] -# Either "memory", "file", "redis", "mysql", "postgres", default is "file" -provider = {{ default "file" .SESSION_PROVIDER }} - -# Provider config options -# memory: not have any config yet -# file: session dir path, is relative to grafana data_path -# redis: config like redis server e.g. `addr=127.0.0.1:6379,pool_size=100,db=grafana` -# mysql: go-sql-driver/mysql dsn config string, e.g. `user:password@tcp(127.0.0.1:3306)/database_name` -# postgres: user=a password=b host=localhost port=5432 dbname=c sslmode=disable -{{ if .SESSION_PROVIDER_CONFIG }} -provider_config = {{ .SESSION_PROVIDER_CONFIG }} -{{ end }} - -# Session cookie name -cookie_name = {{ default "grafana_sess" .SESSION_COOKIE_NAME }} - -# If you use session in https only, default is false -cookie_secure = {{ default "false" .SESSION_COOKIE_SECURE }} - -# Session life time, default is 86400 -session_life_time = {{ default "86400" .SESSION_LIFE_TIME }} - -#################################### Analytics #################################### -[analytics] -# Server reporting, sends usage counters to stats.grafana.org every 24 hours. -# No ip addresses are being tracked, only simple counters to track -# running instances, dashboard and error counts. It is very helpful to us. -# Change this option to false to disable reporting. -reporting_enabled = {{ default "false" .GRAFANA_REPORTING }} - -# Google Analytics universal tracking code, only enabled if you specify an id here -{{ if .GOOGLE_ANALYTICS_UA_ID }} -google_analytics_ua_id = {{ .GOOGLE_ANALYTICS_UA_ID }} -{{ end }} - -#################################### Security #################################### -[security] -# default admin user, created on startup -admin_user = {{ default "admin" .DEFAULT_USER }} - -# default admin password, can be changed before first start of grafana, or in profile settings -admin_password = {{ default "admin" .DEFAULT_USER_PASSWORD }} - -# used for signing -secret_key = {{ default "SW2YcwTIb9zpOOhoPsMm" .SECRET_KEY }} - -# Auto-login remember days -login_remember_days = {{ default "7" .LOGIN_REMEMBER_DAYS }} -cookie_username = {{ default "grafana_user" .COOKIE_USERNAME }} -cookie_remember_name = {{ default "grafana_remember" .COOKIE_REMEMBER_NAME }} - -# disable gravatar profile images -disable_gravatar = {{ default "false" .DISABLE_GRAVATAR }} - -# data source proxy whitelist (ip_or_domain:port seperated by spaces) -{{ if .DATASOURCE_PROXY_WHITELIST }} -data_source_proxy_whitelist = {{ .DATA_SOURCE_PROXY_WHITELIST }} -{{ end }} - -#################################### Users #################################### -[users] -# disable user signup / registration -allow_sign_up = {{ default "true" .ALLOW_SIGN_UP }} - -# Allow non admin users to create organizations -allow_org_create = {{ default "true" .ALLOW_ORG_CREATE }} - -# Set to true to automatically assign new users to the default organization (id 1) -auto_assign_org = {{ default "true" .AUTO_ASSIGN_ORG }} - -# Default role new users will be automatically assigned (if disabled above is set to true) -auto_assign_org_role = {{ default "Viewer" .AUTO_ASSIGN_ORG_ROLE }} - -# Background text for the user field on the login page -login_hint = {{ default "email or username" .LOGIN_HINT }} -#################################### Auth #################################### -[auth] -oauth_auto_login = true - -#################################### Anonymous Auth ########################## -[auth.anonymous] -# enable anonymous access -enabled = {{ default "false" .ANON_ACCESS }} - -# specify organization name that should be used for unauthenticated users -org_name = {{ default "Main Org." .ORG_NAME }} - -# specify role for unauthenticated users -org_role = {{ default "Viewer" .UNAUTHED_USER_ORG_ROLE }} - -#################################### Github Auth ########################## -[auth.github] -{{ if .GITHUB_AUTH }} -enabled = {{ .GITHUB_AUTH }} -allow_sign_up = {{ default "false" .GITHUB_AUTH_ALLOW_SIGN_UP }} -client_id = {{ .GITHUB_AUTH_CLIENT_ID }} -client_secret = {{ .GITHUB_AUTH_CLIENT_SECRET }} -scopes = {{ default "user:email,read:org" .GITHUB_AUTH_SCOPES }} -auth_url = {{ default "https://github.com/login/oauth/authorize" .GITHUB_AUTH_URL }} -token_url = {{ default "https://github.com/login/oauth/access_token" .GITHUB_AUTH_TOKEN_URL }} -api_url = {{ default "https://api.github.com/user" .GITHUB_AUTH_API_URL }} -team_ids = {{ default "" .GITHUB_AUTH_TEAM_IDS }} -allowed_organizations = {{ default "" .GITHUB_AUTH_ALLOWED_ORGS}} -{{ end }} - -#################################### Google Auth ########################## -[auth.google] -{{ if .GOOGLE_AUTH }} -enabled = {{ .GOOGLE_AUTH }} -allow_sign_up = {{ default "false" .GOOGLE_AUTH_ALLOW_SIGN_UP }} -client_id = {{ .GOOGLE_AUTH_CLIENT_ID }} -client_secret = {{ .GOOGLE_AUTH_CLIENT_SECRET }} -scopes = {{ default "https://www.googleapis.com/auth/userinfo.profile ttps://www.googleapis.com/auth/userinfo.email" .GOOGLE_AUTH_SCOPES }} -auth_url = {{ default "https://accounts.google.com/o/oauth2/auth" .GOOGLE_AUTH_URL }} -token_url = {{ default "https://accounts.google.com/o/oauth2/token" .GOOGLE_AUTH_TOKEN_URL }} -api_url = {{ default "https://www.googleapis.com/oauth2/v1/userinfo" .GOOGLE_AUTH_API_URL }} -allowed_domains = {{ default "" .GOOGLE_AUTH_ALLOWED_DOMAINS }} -{{ end }} - -#################################### Auth Proxy ########################## -[auth.proxy] -{{ if .AUTH_PROXY }} -enabled = {{ .AUTH_PROXY }} -header_name = {{ default "X-WEBAUTH-USER" .AUTH_HEADER_NAME }} -header_property = {{ default "username" .AUTH_HEADER_PROPERTY }} -auto_sign_up = {{ default "true" .AUTH_AUTO_SIGN_UP }} -{{ end }} - -#################################### Basic Auth ########################## -[auth.basic] -enabled = {{ default "true" .BASIC_AUTH }} - -#################################### Auth LDAP ########################## -[auth.ldap] -{{ if .LDAP_ENDPOINT }} -enabled = true -config_file = "/usr/share/grafana/ldap.toml" -{{ end }} - -#################################### Generic Oauth ########################## -[auth.generic_oauth] -{{ if .DRYCC_PASSPORT_URL }} -name = OAuth -enabled = true -client_id = {{ .DRYCC_PASSPORT_KEY }} -client_secret = {{ .DRYCC_PASSPORT_SECRET }} -scopes = profile,openid -auth_url = {{ .DRYCC_PASSPORT_URL }}/oauth/authorize/ -token_url = {{ .DRYCC_PASSPORT_URL }}/oauth/token/ -api_url = {{ .DRYCC_PASSPORT_URL }} -allow_sign_up = true -tls_skip_verify_insecure = true -role_attribute_path = (is_superuser && 'Admin') || (is_staff && 'Viewer') || '' -role_attribute_strict = true - -[server] -root_url = {{ .DRYCC_GRAFANA_URL }} - -{{ end }} - -#################################### SMTP / Emailing ########################## -[smtp] -{{ if .SMTP }} -enabled = {{ .SMTP }} -host = {{ .SMTP_HOST }} -user = {{ .SMTP_USER }} -password = {{ .SMTP_PASSWORD }} -cert_file = {{ .SMTP_CERT_FILE }} -key_file = {{ .SMTP_KEY_FILE }} -skip_verify = {{ default "false" .SMTP_SKIP_VERIFY }} -from_address = {{ default "admin@grafana.localhost" .SMTP_FROM_ADDRESS }} -{{ end }} - -[emails] -welcome_email_on_sign_up = {{ default "false" .WELCOME_EMAIL }} - -#################################### Logging ########################## -[log] -# Either "console", "file", default is "console" -# Use comma to separate multiple modes, e.g. "console, file" -mode = {{ default "console" .LOG_MODE }} - -# Buffer length of channel, keep it as it is if you don't know what it is. -buffer_len = {{ default "10000" .BUFFER_LENGTH }} - -# Either "Trace", "Debug", "Info", "Warn", "Error", "Critical", default is "Trace" -level = {{ default "Info" .LOG_LEVEL }} - -# For "console" mode only -[log.console] -level = {{ default "Info" .LOG_LEVEL_CONSOLE }} - -# For "file" mode only -[log.file] -level = {{ default "Info" .LOG_LEVEL_FILE }} -# This enables automated log rotate(switch of following options), default is true -log_rotate = {{ default "true" .LOG_ROTATE }} - -# Max line number of single file, default is 1000000 -max_lines = {{ default "1000000" .LOG_MAX_LINES }} - -# Max size shift of single file, default is 28 means 1 << 28, 256MB -max_lines_shift = {{ default "28" .LOG_MAX_LINES_SHIFT }} - -# Segment log daily, default is true -daily_rotate = {{ default "true" .LOG_DAILY_ROTATE }} - -# Expired days of log file(delete after max days), default is 7 -max_days = {{ default "7" .LOG_MAX_DAYS }} - -#################################### AMPQ Event Publisher ########################## -[event_publisher] -{{ if .AMQP_PUBLISHER }} -enabled = {{ .AMQP_PUBLISHER }} -rabbitmq_url = {{ .AMQP_PUBLISHER_HOST }} -exchange = {{ .AMQP_PUBLISHER_EXCHANGE }} -{{ end }} - -;#################################### Dashboard JSON files ########################## -[dashboards.json] -enabled = {{ default "true" .DASHBOARD_JSON }} -path = {{ default "/usr/share/grafana/dashboards" .DASHBOARD_JSON_PATH }} diff --git a/grafana/rootfs/usr/share/grafana/ldap.toml.tpl b/grafana/rootfs/usr/share/grafana/ldap.toml.tpl deleted file mode 100644 index 1c8b617..0000000 --- a/grafana/rootfs/usr/share/grafana/ldap.toml.tpl +++ /dev/null @@ -1,45 +0,0 @@ -[[servers]] -# Ldap server host (specify multiple hosts space separated) -host = "{{ .LDAP_HOST }}" -# Default port is 389 or 636 if use_ssl = true -port = {{ .LDAP_PORT }} -# Set to true if LDAP server supports TLS -use_ssl = {{ .LDAP_USE_SSL }} -# Set to true if connect LDAP server with STARTTLS pattern (create connection in insecure, then upgrade to secure connection with TLS) -start_tls = true -# set to true if you want to skip SSL cert validation -ssl_skip_verify = true - -# Search user bind dn -bind_dn = "{{ .LDAP_BIND_DN }}" -# Search user bind password -# If the password contains # or ; you have to wrap it with triple quotes. Ex """#password;""" -bind_password = "{{ .LDAP_BIND_PASSWORD }}" - -# User search filter, for example "(cn=%s)" or "(sAMAccountName=%s)" or "(uid=%s)" -# Allow login from email or username, example "(|(sAMAccountName=%s)(userPrincipalName=%s))" -search_filter = "{{ .LDAP_USER_FILTER }}" - -# An array of base dns to search through -search_base_dns = ["{{ .LDAP_USER_BASEDN }}"] - -# Specify names of the LDAP attributes your LDAP uses -[servers.attributes] -name = "givenName" -surname = "sn" -username = "cn" -member_of = "memberOf" -email = "mail" - -[[servers.group_mappings]] -group_dn = "{{ .LDAP_ADMIN_GROUP }}" -org_role = "Admin" -grafana_admin = true # Available in Grafana v5.3 and above - -[[servers.group_mappings]] -group_dn = "{{ .LDAP_EDITOR_GROUP }}" -org_role = "Editor" - -[[servers.group_mappings]] -group_dn = "{{ .LDAP_VIEWER_GROUP }}" -org_role = "Viewer" diff --git a/grafana/rootfs/usr/share/grafana/start-grafana b/grafana/rootfs/usr/share/grafana/start-grafana deleted file mode 100755 index 4987128..0000000 --- a/grafana/rootfs/usr/share/grafana/start-grafana +++ /dev/null @@ -1,113 +0,0 @@ -#!/bin/bash - -# Some of the contents of this file came from here - https://github.com/kubernetes/heapster/blob/ed5baadf04ea9f8e48fc7d44dad63a63af34ff9b/grafana/run.sh -HEADER_CONTENT_TYPE="Content-Type: application/json" -HEADER_ACCEPT="Accept: application/json" - -GRAFANA_USER=${DEFAULT_USER:-admin} -GRAFANA_PASSWD=${DEFAULT_USER_PASSWORD:-admin} -GRAFANA_PORT=${BIND_PORT:-3000} - -INFLUXDB_HOST=${DRYCC_INFLUXDB_SERVICE_HOST:-"drycc-influxdb.drycc.svc.${KUBERNETES_CLUSTER_DOMAIN}"} -INFLUXDB_PORT=${DRYCC_INFLUXDB_SERVICE_PORT:-80} -INFLUXDB_BUCKET=${INFLUXDB_BUCKET:-kubernetes} -INFLUXDB_ORG=${INFLUXDB_ORG:-drycc} -INFLUXDB_VERSION=${INFLUXDB_VERSION:-Flux} - -DASHBOARD_LOCATION=${DASHBOARD_JSON_PATH:-"/usr/share/grafana/api/dashboards"} -# Create the dashboards directory -mkdir /usr/share/grafana/dashboards - -echo "Building grafana.ini!" -envtpl > grafana.ini < grafana.ini.tpl -echo "Finished building grafana config..." -echo "###########################################" -echo "###########################################" -cat /usr/share/grafana/grafana.ini -echo "###########################################" -echo "###########################################" - -if [ -n "${LDAP_ENDPOINT}" ]; then - echo "Building ldap.toml!" - OLD_IFS="$IFS" - IFS=':/' read -ra ldap_url_array <<< "$LDAP_ENDPOINT" - IFS="$OLD_IFS" - ldap_schema=${ldap_url_array[0]} - if [ "${ldap_schema}" == "ldaps" ]; then - export LDAP_USE_SSL=true - else - export LDAP_USE_SSL=false - fi - export LDAP_HOST=${ldap_url_array[3]//\/\//} - export LDAP_PORT=${ldap_url_array[4]} - envtpl > ldap.toml < ldap.toml.tpl - echo "Finished building ldap config..." - echo "###########################################" - echo "###########################################" - cat /usr/share/grafana/ldap.toml - echo "###########################################" - echo "###########################################" -fi - -GF_PATHS_PLUGINS=${PLUGINS_PATH:-"/var/lib/grafana/plugins"} -if [ -n "${GF_INSTALL_PLUGINS}" ]; then - echo "Installing Grafana plugins..." - OLDIFS=$IFS - IFS=',' - for plugin in ${GF_INSTALL_PLUGINS}; do - echo "Installing ${plugin} ..." - grafana-cli --pluginsDir "${GF_PATHS_PLUGINS}" plugins install "${plugin}" - echo "" - echo "Done installing ${plugin}" - done - IFS=$OLDIFS -fi - -set -m -echo "Starting Grafana in the background" -exec grafana-server -config /usr/share/grafana/grafana.ini -homepath /opt/drycc/grafana & -echo "Waiting for Grafana to come up..." -until curl -q --fail --output /dev/null --silent "http://${GRAFANA_USER}:${GRAFANA_PASSWD}@localhost:${GRAFANA_PORT}/api/org"; do - printf "." - sleep 2 -done -echo "Grafana is up and running." - -# Allow access to dashboards without having to log in -export GF_AUTH_ANONYMOUS_ENABLED=true -export GF_SERVER_HTTP_PORT=${GRAFANA_PORT} - -BACKEND_ACCESS_MODE=${BACKEND_ACCESS_MODE:-proxy} -INFLUXDB_URL=${INFLUXDB_URL} -if [ -n "$INFLUXDB_URL" ]; then - echo "Influxdb service URL is provided." -else - INFLUXDB_URL="http://${INFLUXDB_HOST}:${INFLUXDB_PORT}" -fi - -echo "Using the following URL for InfluxDB: ${INFLUXDB_URL}" -echo "Using the following backend access mode for InfluxDB: ${BACKEND_ACCESS_MODE}" - -echo "Creating default influxdb datasource..." -curl -i -XPOST -H "${HEADER_ACCEPT}" -H "${HEADER_CONTENT_TYPE}" "http://${GRAFANA_USER}:${GRAFANA_PASSWD}@localhost:${GRAFANA_PORT}/api/datasources" -d ' -{ - "name": "influxdb-datasource", - "type": "influxdb", - "access": "'"${BACKEND_ACCESS_MODE}"'", - "isDefault": true, - "url": "'"${INFLUXDB_URL}"'", - "secureJsonData": {"token": "'"${INFLUXDB_TOKEN}"'"}, - "jsonData": {"version": "'"${INFLUXDB_VERSION}"'", "httpMode": "POST", "organization": "'"${INFLUXDB_ORG}"'", "defaultBucket": "'"${INFLUXDB_BUCKET}"'"} -}' - -echo "" -echo "Importing default dashboards..." -for filename in "${DASHBOARD_LOCATION}"/*.json; do - echo "Importing ${filename} ..." - curl -i -XPOST --data "@${filename}" -H "${HEADER_ACCEPT}" -H "${HEADER_CONTENT_TYPE}" "http://${GRAFANA_USER}:${GRAFANA_PASSWD}@localhost:${GRAFANA_PORT}/api/dashboards/db" - echo "" - echo "Done importing ${filename}" -done - -echo "Bringing Grafana back to the foreground" -fg diff --git a/includes.mk b/includes.mk index ea2927b..fde07fe 100644 --- a/includes.mk +++ b/includes.mk @@ -1,7 +1,7 @@ -.PHONY: check-docker -check-docker: - @if [ -z $$(which docker) ]; then \ - echo "Missing \`docker\` client which is required for development"; \ +.PHONY: check-podman +check-podman: + @if [ -z $$(which podman) ]; then \ + echo "Missing \`podman\` client which is required for development"; \ exit 2; \ fi diff --git a/rootfs/Dockerfile b/rootfs/Dockerfile new file mode 100644 index 0000000..66d795e --- /dev/null +++ b/rootfs/Dockerfile @@ -0,0 +1,52 @@ +ARG CODENAME +FROM registry.drycc.cc/drycc/base:${CODENAME} + +ARG DRYCC_UID=1001 \ + DRYCC_GID=1001 \ + DRYCC_HOME_DIR=/usr/share/grafana \ + GRAFANA_VERSION="13.0.1" \ + JQ_VERSION="1.7.1" \ + CADDY_VERSION="2.11.2" \ + PYTHON_VERSION="3.14" + +RUN groupadd drycc --gid ${DRYCC_GID} \ + && useradd drycc -u ${DRYCC_UID} -g ${DRYCC_GID} -s /bin/bash -m -d ${DRYCC_HOME_DIR} + +COPY . / + +RUN install-packages unzip fontconfig \ + && install-stack grafana ${GRAFANA_VERSION} \ + && install-stack jq ${JQ_VERSION} \ + && install-stack caddy $CADDY_VERSION \ + && install-stack python $PYTHON_VERSION && . init-stack \ + && python3 -m venv ${DRYCC_HOME_DIR}/venv \ + && source ${DRYCC_HOME_DIR}/venv/bin/activate \ + && pip3 install --disable-pip-version-check --no-cache-dir -r ${DRYCC_HOME_DIR}/oauth2/requirements.txt \ + && chown -R ${DRYCC_UID}:${DRYCC_GID} ${DRYCC_HOME_DIR} \ + # set env + && echo "source ${DRYCC_HOME_DIR}/venv/bin/activate" >> /opt/drycc/python/profile.d/python.sh \ + # grafana plugins + && mkdir ${DRYCC_HOME_DIR}/plugins \ + && grafana cli --pluginsDir=${DRYCC_HOME_DIR}/plugins --homepath=/opt/drycc/grafana plugins install grafana-llm-app \ + && grafana cli --pluginsDir=${DRYCC_HOME_DIR}/plugins --homepath=/opt/drycc/grafana plugins install quickwit-quickwit-datasource \ + # clean + && rm -rf \ + /usr/share/doc \ + /usr/share/man \ + /usr/share/info \ + /usr/share/locale \ + /var/lib/apt/lists/* \ + /var/log/* \ + /var/cache/debconf/* \ + /etc/systemd \ + /lib/lsb \ + /lib/udev \ + /usr/lib/`echo $(uname -m)`-linux-gnu/gconv/IBM* \ + /usr/lib/`echo $(uname -m)`-linux-gnu/gconv/EBC* \ + && mkdir -p /usr/share/man/man{1..8} \ + && mkdir /var/lib/grafana -p \ + && chown -R ${DRYCC_UID}:${DRYCC_GID} /opt/drycc/grafana /usr/share/grafana /var/lib/grafana + +USER ${DRYCC_UID} +WORKDIR ${DRYCC_HOME_DIR} +EXPOSE 3000 diff --git a/rootfs/usr/share/grafana/dashboards/drycc_component_health.json b/rootfs/usr/share/grafana/dashboards/drycc_component_health.json new file mode 100644 index 0000000..34db95a --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/drycc_component_health.json @@ -0,0 +1,2770 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "target": { + "limit": 100, + "matchAny": false, + "tags": [], + "type": "dashboard" + }, + "type": "dashboard" + } + ] + }, + "editable": false, + "fiscalYearStartMonth": 0, + "graphTooltip": 0, + "id": null, + "links": [], + "liveNow": false, + "panels": [ + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 33, + "panels": [], + "targets": [ + { + "refId": "A" + } + ], + "title": "SUM", + "type": "row" + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 1 + }, + "hiddenSeries": false, + "id": 31, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_namespace", + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",container!=\"\",namespace=\"drycc\"}[1m])) * 1000", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["namespace"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Total CPU Usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "Millicores", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 1 + }, + "hiddenSeries": false, + "id": 32, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_namespace", + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",container!=\"\",namespace=\"drycc\"})", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["namespace"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Total Memory Usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 8 + }, + "id": 52, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 2 + }, + "hiddenSeries": false, + "id": 54, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",container=~\"^drycc-passport.*$\"}[1m])) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "hide": false, + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "Millicores", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 2 + }, + "hiddenSeries": false, + "id": 56, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-passport.*$\"}) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Memory", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "refId": "A" + } + ], + "title": "PASSPORT", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 9 + }, + "id": 34, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 3 + }, + "hiddenSeries": false, + "id": 1, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-controller.*$\"}[1m])) by (pod) * 1000", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "hide": false, + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "Millicores", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 3 + }, + "hiddenSeries": false, + "id": 2, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-controller.*$\"}) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Memory", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "refId": "A" + } + ], + "title": "CONTROLLER", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 10 + }, + "id": 35, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 4 + }, + "hiddenSeries": false, + "id": 3, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-builder.*$\"}[1m])) by (pod) * 1000", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "Millicores", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 4 + }, + "hiddenSeries": false, + "id": 4, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-builder.*$\"}) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Memory", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "refId": "A" + } + ], + "title": "BUILDER", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 11 + }, + "id": 36, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 12 + }, + "hiddenSeries": false, + "id": 6, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-database.*$\"}[1m])) by (pod) * 1000", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "Millicores", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 12 + }, + "hiddenSeries": false, + "id": 5, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-database.*$\"}) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Memory", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "refId": "A" + } + ], + "title": "DATABASE", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 12 + }, + "id": 37, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 13 + }, + "hiddenSeries": false, + "id": 19, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-logger-fluentbit.*$\"}[1m])) by (pod) * 1000", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "Millicores", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 13 + }, + "hiddenSeries": false, + "id": 20, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-logger-fluentbit.*$\"}) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Memory", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "refId": "A" + } + ], + "title": "FLUENTBIT", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 13 + }, + "id": 38, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": null, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 7 + }, + "hiddenSeries": false, + "id": 25, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "datasource": null, + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-grafana.*$\"}[1m])) by (pod) * 1000", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "Millicores", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": null, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 7 + }, + "hiddenSeries": false, + "id": 26, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "datasource": null, + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-grafana.*$\"}) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Memory", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "refId": "A" + } + ], + "title": "GRAFANA", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 15 + }, + "id": 40, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 16 + }, + "hiddenSeries": false, + "id": 7, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",namespace=\"drycc\",container=\"drycc-logger\"}[1m])) by (pod) * 1000", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "Millicores", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 16 + }, + "hiddenSeries": false, + "id": 8, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",namespace=\"drycc\",container=\"drycc-logger\"}) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Memory", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "refId": "A" + } + ], + "title": "LOGGER", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 16 + }, + "id": 41, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 17 + }, + "hiddenSeries": false, + "id": 9, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-storage.*$\"}[1m])) by (pod) * 1000", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "Millicores", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 17 + }, + "hiddenSeries": false, + "id": 10, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-storage.*$\"}) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Memory", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "refId": "A" + } + ], + "title": "STORAGE", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 20 + }, + "id": 43, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 21 + }, + "hiddenSeries": false, + "id": 27, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-valkey.*$\"}[1m])) by (pod) * 1000", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 21 + }, + "hiddenSeries": false, + "id": 28, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-valkey.*$\"}) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Memory", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "refId": "A" + } + ], + "title": "VALKEY", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 21 + }, + "id": 44, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 22 + }, + "hiddenSeries": false, + "id": 11, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-registry.*$\"}[1m])) by (pod) * 1000", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "Millicores", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 22 + }, + "hiddenSeries": false, + "id": 12, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",namespace=\"drycc\",container=\"drycc-registry\"}) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Memory", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "refId": "A" + } + ], + "title": "REGISTRY", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 22 + }, + "id": 45, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 23 + }, + "hiddenSeries": false, + "id": 16, + "interval": "", + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-registry-proxy.*$\"}[1m])) by (pod) * 1000", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "Millicores", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 23 + }, + "hiddenSeries": false, + "id": 15, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.1.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "$tag_pod_name", + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",namespace=\"drycc\",container=~\"^drycc-registry-proxy.*$\"}) by (pod)", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["pod_name"], + "type": "tag" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "legendFormat": "__auto", + "orderByTime": "ASC", + "policy": "default", + "range": true, + "refId": "A", + "resultFormat": "time_series" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Memory", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "refId": "A" + } + ], + "title": "REGISTRY-PROXY", + "type": "row" + } + ], + "refresh": "5s", + "schemaVersion": 37, + "style": "dark", + "tags": ["drycc"], + "templating": { + "list": [] + }, + "time": { + "from": "now-3h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": ["5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d"] + }, + "timezone": "browser", + "title": "Drycc Component Health", + "uid": "2ySBrTTMz", + "version": 1, + "weekStart": "" +} diff --git a/rootfs/usr/share/grafana/dashboards/drycc_database.json b/rootfs/usr/share/grafana/dashboards/drycc_database.json new file mode 100644 index 0000000..db2083b --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/drycc_database.json @@ -0,0 +1,1327 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": "-- Grafana --", + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "description": "A quickstart to setup the Prometheus PostgreSQL Exporter with preconfigured dashboards, alerting rules, and recording rules.", + "editable": false, + "gnetId": 14114, + "graphTooltip": 0, + "id": 1, + "iteration": 1603191461722, + "links": [], + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 20, + "x": 0, + "y": 0 + }, + "hiddenSeries": false, + "id": 1, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": false, + "max": true, + "min": true, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "7.2.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "fetched", + "dsType": "prometheus", + "expr": "sum(irate(pg_stat_database_tup_fetched{datname=~\"$db\",instance=~\"$instance\"}[5m]))", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "fetched", + "measurement": "postgresql", + "policy": "default", + "refId": "A", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["tup_fetched"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": ["10s"], + "type": "non_negative_derivative" + } + ] + ], + "step": 120, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + }, + { + "alias": "fetched", + "dsType": "prometheus", + "expr": "sum(irate(pg_stat_database_tup_returned{datname=~\"$db\",instance=~\"$instance\"}[5m]))", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "returned", + "measurement": "postgresql", + "policy": "default", + "refId": "B", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["tup_fetched"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": ["10s"], + "type": "non_negative_derivative" + } + ] + ], + "step": 120, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + }, + { + "alias": "fetched", + "dsType": "prometheus", + "expr": "sum(irate(pg_stat_database_tup_inserted{datname=~\"$db\",instance=~\"$instance\"}[5m]))", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "inserted", + "measurement": "postgresql", + "policy": "default", + "refId": "C", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["tup_fetched"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": ["10s"], + "type": "non_negative_derivative" + } + ] + ], + "step": 120, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + }, + { + "alias": "fetched", + "dsType": "prometheus", + "expr": "sum(irate(pg_stat_database_tup_updated{datname=~\"$db\",instance=~\"$instance\"}[5m]))", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "updated", + "measurement": "postgresql", + "policy": "default", + "refId": "D", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["tup_fetched"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": ["10s"], + "type": "non_negative_derivative" + } + ] + ], + "step": 120, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + }, + { + "alias": "fetched", + "dsType": "prometheus", + "expr": "sum(irate(pg_stat_database_tup_deleted{datname=~\"$db\",instance=~\"$instance\"}[5m]))", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "deleted", + "measurement": "postgresql", + "policy": "default", + "refId": "E", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["tup_fetched"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": ["10s"], + "type": "non_negative_derivative" + } + ] + ], + "step": 120, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Rows", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(245, 54, 54, 0.9)", + "rgba(237, 129, 40, 0.89)", + "rgba(50, 172, 45, 0.97)" + ], + "decimals": 0, + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "format": "none", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 20, + "y": 0 + }, + "height": "55px", + "id": 11, + "interval": null, + "isNew": true, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": true, + "lineColor": "rgb(31, 120, 193)", + "show": true + }, + "tableColumn": "", + "targets": [ + { + "dsType": "prometheus", + "expr": "sum(irate(pg_stat_database_xact_commit{datname=~\"$db\",instance=~\"$instance\"}[5m])) + sum(irate(pg_stat_database_xact_rollback{datname=~\"$db\",instance=~\"$instance\"}[5m]))", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "measurement": "postgresql", + "policy": "default", + "refId": "A", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["xact_commit"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": ["10s"], + "type": "non_negative_derivative" + } + ] + ], + "step": 1800, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + } + ], + "thresholds": "", + "title": "QPS", + "transparent": true, + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "avg" + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "decimals": 1, + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 7 + }, + "hiddenSeries": false, + "id": 2, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": false, + "hideZero": true, + "max": true, + "min": true, + "rightSide": false, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "7.2.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "Buffers Allocated", + "dsType": "prometheus", + "expr": "irate(pg_stat_bgwriter_buffers_alloc{instance=~'$instance'}[5m])", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "buffers_alloc", + "measurement": "postgresql", + "policy": "default", + "refId": "A", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["buffers_alloc"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": [], + "type": "difference" + } + ] + ], + "step": 240, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + }, + { + "alias": "Buffers Allocated", + "dsType": "prometheus", + "expr": "irate(pg_stat_bgwriter_buffers_backend_fsync{instance=~'$instance'}[5m])", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "buffers_backend_fsync", + "measurement": "postgresql", + "policy": "default", + "refId": "B", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["buffers_alloc"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": [], + "type": "difference" + } + ] + ], + "step": 240, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + }, + { + "alias": "Buffers Allocated", + "dsType": "prometheus", + "expr": "irate(pg_stat_bgwriter_buffers_backend{instance=~'$instance'}[5m])", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "buffers_backend", + "measurement": "postgresql", + "policy": "default", + "refId": "C", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["buffers_alloc"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": [], + "type": "difference" + } + ] + ], + "step": 240, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + }, + { + "alias": "Buffers Allocated", + "dsType": "prometheus", + "expr": "irate(pg_stat_bgwriter_buffers_clean{instance=~'$instance'}[5m])", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "buffers_clean", + "measurement": "postgresql", + "policy": "default", + "refId": "D", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["buffers_alloc"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": [], + "type": "difference" + } + ] + ], + "step": 240, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + }, + { + "alias": "Buffers Allocated", + "dsType": "prometheus", + "expr": "irate(pg_stat_bgwriter_buffers_checkpoint{instance=~'$instance'}[5m])", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "buffers_checkpoint", + "measurement": "postgresql", + "policy": "default", + "refId": "E", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["buffers_alloc"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": [], + "type": "difference" + } + ] + ], + "step": 240, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Buffers", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 7 + }, + "hiddenSeries": false, + "id": 3, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "7.2.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "alias": "conflicts", + "dsType": "prometheus", + "expr": "sum(rate(pg_stat_database_deadlocks{datname=~\"$db\",instance=~\"$instance\"}[5m]))", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "deadlocks", + "measurement": "postgresql", + "policy": "default", + "refId": "A", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["conflicts"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": [], + "type": "difference" + } + ] + ], + "step": 240, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + }, + { + "alias": "deadlocks", + "dsType": "prometheus", + "expr": "sum(rate(pg_stat_database_conflicts{datname=~\"$db\",instance=~\"$instance\"}[5m]))", + "format": "time_series", + "groupBy": [ + { + "params": ["$interval"], + "type": "time" + }, + { + "params": ["null"], + "type": "fill" + } + ], + "intervalFactor": 2, + "legendFormat": "conflicts", + "measurement": "postgresql", + "policy": "default", + "refId": "B", + "resultFormat": "time_series", + "select": [ + [ + { + "params": ["deadlocks"], + "type": "field" + }, + { + "params": [], + "type": "mean" + }, + { + "params": [], + "type": "difference" + } + ] + ], + "step": 240, + "tags": [ + { + "key": "instance", + "operator": "=~", + "value": "/^$instance$/" + } + ] + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Conflicts/Deadlocks", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 14 + }, + "hiddenSeries": false, + "id": 12, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": true, + "pluginVersion": "7.2.1", + "pointradius": 1, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(pg_stat_database_blks_hit{datname=~\"$db\",instance=~\"$instance\"}) / (sum(pg_stat_database_blks_hit{datname=~\"$db\",instance=~\"$instance\"}) + sum(pg_stat_database_blks_read{datname=~\"$db\",instance=~\"$instance\"}))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "cache hit rate", + "refId": "A", + "step": 240 + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Cache hit ratio", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 14 + }, + "hiddenSeries": false, + "id": 13, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "7.2.1", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "pg_stat_database_numbackends{datname=~\"$db\",instance=~\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{__name__}}", + "refId": "A", + "step": 240 + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Number of active connections", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + } + ], + "refresh": false, + "schemaVersion": 26, + "style": "dark", + "tags": ["postgres"], + "templating": { + "list": [ + { + "allValue": ".*", + "current": { + "selected": false, + "text": "All", + "value": "$__all" + }, + "definition": "", + "hide": 0, + "includeAll": true, + "label": null, + "multi": false, + "name": "instance", + "options": [], + "query": "label_values(up{job=~\"postgres.*\"},instance)", + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "tagValuesQuery": "", + "tags": [], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": ".*", + "current": { + "selected": false, + "text": "All", + "value": "$__all" + }, + "definition": "label_values(pg_stat_database_tup_fetched{instance=~\"$instance\",datname!~\"template.*|postgres\"},datname)", + "hide": 0, + "includeAll": true, + "label": "db", + "multi": false, + "name": "db", + "options": [], + "query": "label_values(pg_stat_database_tup_fetched{instance=~\"$instance\",datname!~\"template.*|postgres\"},datname)", + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "tagValuesQuery": "", + "tags": [], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "current": { + "selected": false, + "text": "Postgres Overview", + "value": "Postgres Overview" + }, + "hide": 0, + "includeAll": false, + "label": "datasource", + "multi": false, + "name": "datasource", + "options": [], + "query": "prometheus", + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "type": "datasource" + }, + { + "allValue": null, + "current": { + "selected": true, + "text": "postgres", + "value": "postgres" + }, + "definition": "label_values(pg_up, job)", + "hide": 0, + "includeAll": false, + "label": "job", + "multi": false, + "name": "job", + "options": [ + { + "selected": true, + "text": "postgres", + "value": "postgres" + } + ], + "query": "label_values(pg_up, job)", + "refresh": 0, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "tagValuesQuery": "", + "tags": [], + "tagsQuery": "", + "type": "query", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": ["5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d"] + }, + "timezone": "browser", + "title": "PostgreSQL Exporter Quickstart and Dashboard", + "uid": "wGgaPlciz", + "version": 5 +} diff --git a/rootfs/usr/share/grafana/dashboards/drycc_quickwit.json b/rootfs/usr/share/grafana/dashboards/drycc_quickwit.json new file mode 100644 index 0000000..e707c56 --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/drycc_quickwit.json @@ -0,0 +1,1286 @@ +{ + "__inputs": [ + { + "name": "datasource", + "label": "Prometheus", + "description": "", + "type": "datasource", + "pluginId": "prometheus", + "pluginName": "Prometheus" + } + ], + "__elements": {}, + "__requires": [ + { + "type": "grafana", + "id": "grafana", + "name": "Grafana", + "version": "10.0.1" + }, + { + "type": "datasource", + "id": "prometheus", + "name": "Prometheus", + "version": "1.0.0" + }, + { + "type": "panel", + "id": "stat", + "name": "Stat", + "version": "" + }, + { + "type": "panel", + "id": "timeseries", + "name": "Time series", + "version": "" + } + ], + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": { + "type": "grafana", + "uid": "-- Grafana --" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "target": { + "limit": 100, + "matchAny": false, + "tags": [], + "type": "dashboard" + }, + "type": "dashboard" + } + ] + }, + "description": "Dashboard to monitor Quickwit indexers.", + "editable": false, + "fiscalYearStartMonth": 0, + "graphTooltip": 0, + "id": null, + "links": [], + "liveNow": false, + "panels": [ + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "Bps" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 6, + "x": 0, + "y": 0 + }, + "id": 10, + "options": { + "colorMode": "value", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "10.0.1", + "targets": [ + { + "editorMode": "builder", + "expr": "sum by(docs_processed_status) (rate(quickwit_indexing_processed_bytes{index=~\"$index\", namespace=\"$namespace\", instance=~\"$instance\", pod=~\"$pod\"}[$quickwit_rate_interval]))", + "legendFormat": "{{docs_processed_status}}", + "range": true, + "refId": "A" + } + ], + "title": "Indexing throughput", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "suffix: docs/s" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 6, + "x": 6, + "y": 0 + }, + "id": 11, + "options": { + "colorMode": "value", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "10.0.1", + "targets": [ + { + "editorMode": "builder", + "expr": "sum by(docs_processed_status) (rate(quickwit_indexing_processed_docs_total{index=~\"$index\", namespace=\"$namespace\", pod=~\"$pod\"}[$quickwit_rate_interval]))", + "legendFormat": "{{docs_processed_status}}", + "range": true, + "refId": "A" + } + ], + "title": "Documents throughput", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "left", + "axisSoftMax": -4, + "axisSoftMin": 8, + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 8, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "Bps" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "valid" + }, + "properties": [ + { + "id": "color", + "value": { + "mode": "continuous-GrYlRd", + "seriesBy": "last" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "parsing_error" + }, + "properties": [ + { + "id": "color", + "value": { + "mode": "continuous-RdYlGr" + } + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 0 + }, + "id": 2, + "options": { + "legend": { + "calcs": [ + "min", + "max", + "mean" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "editorMode": "builder", + "exemplar": false, + "expr": "sum by(docs_processed_status, index) (rate(quickwit_indexing_processed_bytes{index=~\"$index\", namespace=\"$namespace\", pod=~\"$pod\", instance=~\"$instance\"}[$quickwit_rate_interval]))", + "format": "time_series", + "instant": false, + "interval": "", + "legendFormat": "{{docs_processed_status}}-{{index}}", + "range": true, + "refId": "Indexing bytes rate" + } + ], + "title": "Indexing throughput", + "transformations": [], + "type": "timeseries" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 1, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "Bps" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 8 + }, + "id": 6, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "editorMode": "builder", + "exemplar": false, + "expr": "rate(quickwit_write_bytes{index=~\"$index\", pod=~\"$pod\", namespace=\"$namespace\"}[$quickwit_rate_interval])", + "hide": false, + "instant": false, + "legendFormat": "{{component}} {{index}} pod:{{pod}}", + "range": true, + "refId": "Writes" + } + ], + "title": "Writes rate", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 12, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "decbytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 8 + }, + "id": 8, + "options": { + "legend": { + "calcs": [ + "min", + "max", + "mean" + ], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "9.2.1", + "targets": [ + { + "editorMode": "builder", + "expr": "quickwit_allocated_num_bytes{namespace=\"$namespace\", pod=~\"$pod\"}", + "legendFormat": "allocated-{{pod}}", + "range": true, + "refId": "A" + }, + { + "editorMode": "builder", + "expr": "process_resident_memory_bytes{namespace=\"$namespace\", pod=~\"$pod\"}", + "hide": false, + "legendFormat": "rss-{{pod}}", + "range": true, + "refId": "B" + } + ], + "title": "Memory consumption (allocated and RSS)", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 14, + "gradientMode": "hue", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineStyle": { + "fill": "solid" + }, + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 1, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 16 + }, + "id": 13, + "options": { + "legend": { + "calcs": [ + "min", + "max", + "mean" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "editorMode": "builder", + "expr": "quickwit_indexing_ongoing_merge_operations{index=~\"$index\", namespace=\"$namespace\", instance=~\"$instance\", pod=~\"$pod\"}", + "hide": false, + "legendFormat": "{{index}} {{pod}}", + "range": true, + "refId": "Processed docs" + } + ], + "title": "Ongoing merge operations", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 7, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineStyle": { + "fill": "solid" + }, + "lineWidth": 1, + "pointSize": 4, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 1, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "Bps" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Upload bytes / sec" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 16 + }, + "id": 4, + "options": { + "legend": { + "calcs": [ + "min", + "max", + "mean" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "editorMode": "builder", + "expr": "sum by(pod) (rate(quickwit_storage_object_storage_download_num_bytes{namespace=\"$namespace\", instance=~\"$instance\", pod=~\"$pod\"}[$quickwit_rate_interval]))", + "legendFormat": "Download bytes / sec - {{pod}}", + "range": true, + "refId": "Download" + }, + { + "editorMode": "builder", + "expr": "sum by(pod) (rate(quickwit_storage_object_storage_upload_num_bytes{namespace=\"$namespace\", pod=~\"$pod\", instance=~\"$instance\"}[$__rate_interval]))", + "hide": false, + "legendFormat": "Upload bytes / sec - {{pod}}", + "range": true, + "refId": "Upload" + } + ], + "title": "Object storage transfer rate", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 14, + "gradientMode": "hue", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineStyle": { + "fill": "solid" + }, + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 1, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 24 + }, + "id": 5, + "options": { + "legend": { + "calcs": [ + "min", + "max", + "mean" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "editorMode": "builder", + "expr": "sum by(docs_processed_status, index) (rate(quickwit_indexing_processed_docs_total{index=~\"$index\", namespace=\"$namespace\", instance=~\"$instance\", pod=~\"$pod\"}[$quickwit_rate_interval]))", + "hide": false, + "legendFormat": "{{docs_processed_status}}-{{index}}", + "range": true, + "refId": "Processed docs" + } + ], + "title": "Indexed documents rate", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 7, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineStyle": { + "fill": "solid" + }, + "lineWidth": 1, + "pointSize": 4, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 1, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Upload bytes / sec" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 24 + }, + "id": 14, + "options": { + "legend": { + "calcs": [ + "min", + "max", + "mean" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "editorMode": "builder", + "expr": "sum(rate(quickwit_storage_object_storage_gets_total{namespace=\"$namespace\", instance=~\"$instance\", pod=~\"$pod\"}[$quickwit_rate_interval]))", + "legendFormat": "GET req/sec", + "range": true, + "refId": "Download" + }, + { + "editorMode": "builder", + "expr": "sum(rate(quickwit_storage_object_storage_puts_total{namespace=\"$namespace\", pod=~\"$pod\", instance=~\"$instance\"}[$__rate_interval]))", + "hide": false, + "legendFormat": "PUT req/sec", + "range": true, + "refId": "Upload" + } + ], + "title": "Requests on object storage", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "µs" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 32 + }, + "id": 15, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "editorMode": "builder", + "expr": "histogram_quantile(0.75, rate(quickwit_cli_thread_unpark_duration_microseconds_bucket{namespace=\"$namespace\", pod=~\"$pod\"}[5m]))", + "instant": false, + "legendFormat": "{{pod}}", + "range": true, + "refId": "A" + } + ], + "title": "Thread unpark duration", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 12, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "µs" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 32 + }, + "id": 12, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "9.2.1", + "targets": [ + { + "editorMode": "builder", + "expr": "rate(quickwit_indexing_backpressure_micros{namespace=\"$namespace\", pod=~\"$pod\"}[$__rate_interval])", + "legendFormat": "{{actor_name}}-{{pod}}", + "range": true, + "refId": "A" + } + ], + "title": "Backpressure", + "type": "timeseries" + } + ], + "refresh": "10s", + "schemaVersion": 38, + "style": "dark", + "tags": [ + "quickwit", + "indexer" + ], + "templating": { + "list": [ + { + "current": { + "selected": false, + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "includeAll": false, + "label": "Datasource", + "multi": false, + "name": "datasource", + "options": [], + "query": "prometheus", + "queryValue": "", + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "type": "datasource" + }, + { + "current": {}, + "definition": "label_values(quickwit_write_bytes,namespace)", + "hide": 0, + "includeAll": false, + "label": "Namespace", + "multi": false, + "name": "namespace", + "options": [], + "query": { + "query": "label_values(quickwit_write_bytes,namespace)", + "refId": "StandardVariableQuery" + }, + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "type": "query" + }, + { + "current": {}, + "definition": "label_values(quickwit_write_bytes,index)", + "hide": 0, + "includeAll": true, + "label": "Index", + "multi": false, + "name": "index", + "options": [], + "query": { + "query": "label_values(quickwit_write_bytes,index)", + "refId": "StandardVariableQuery" + }, + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "type": "query" + }, + { + "current": {}, + "definition": "label_values(quickwit_write_bytes,pod)", + "hide": 0, + "includeAll": true, + "label": "Pod", + "multi": true, + "name": "pod", + "options": [], + "query": { + "query": "label_values(quickwit_write_bytes,pod)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "type": "query" + }, + { + "current": {}, + "definition": "label_values(quickwit_write_bytes, instance)", + "hide": 0, + "includeAll": true, + "label": "Instance", + "multi": false, + "name": "instance", + "options": [], + "query": { + "query": "label_values(quickwit_write_bytes, instance)", + "refId": "StandardVariableQuery" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "type": "query" + }, + { + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "selected": false, + "text": "30s", + "value": "30s" + }, + "description": "", + "hide": 0, + "label": "Rate interval", + "name": "quickwit_rate_interval", + "options": [ + { + "selected": false, + "text": "10s", + "value": "10s" + }, + { + "selected": true, + "text": "30s", + "value": "30s" + }, + { + "selected": false, + "text": "1m", + "value": "1m" + }, + { + "selected": false, + "text": "10m", + "value": "10m" + }, + { + "selected": false, + "text": "30m", + "value": "30m" + }, + { + "selected": false, + "text": "1h", + "value": "1h" + }, + { + "selected": false, + "text": "6h", + "value": "6h" + }, + { + "selected": false, + "text": "12h", + "value": "12h" + }, + { + "selected": false, + "text": "1d", + "value": "1d" + }, + { + "selected": false, + "text": "7d", + "value": "7d" + }, + { + "selected": false, + "text": "14d", + "value": "14d" + }, + { + "selected": false, + "text": "30d", + "value": "30d" + } + ], + "query": "10s,30s,1m,10m,30m,1h,6h,12h,1d,7d,14d,30d", + "queryValue": "", + "refresh": 2, + "skipUrlSync": false, + "type": "interval" + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": {}, + "timezone": "utc", + "title": "Quickwit Indexers", + "uid": "quickwit-indexers", + "version": 50, + "weekStart": "", + "gnetId": 20214 +} \ No newline at end of file diff --git a/rootfs/usr/share/grafana/dashboards/drycc_registry.json b/rootfs/usr/share/grafana/dashboards/drycc_registry.json new file mode 100644 index 0000000..d858dc1 --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/drycc_registry.json @@ -0,0 +1,1419 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": "-- Grafana --", + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "editable": false, + "gnetId": 9621, + "graphTooltip": 0, + "id": null, + "iteration": 1547215355938, + "links": [], + "panels": [ + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 22, + "panels": [], + "title": "General Counters, CPU, Memory and File Descriptor Stats", + "type": "row" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": ["#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a"], + "description": "start time of the process", + "format": "dateTimeFromNow", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 0, + "y": 1 + }, + "id": 4, + "interval": null, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "110%", + "prefix": "", + "prefixFontSize": "110%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "process_start_time_seconds{app=\"$app\", instance=\"$instance\"} * 1000", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "", + "title": "Start Time", + "type": "singlestat", + "valueFontSize": "110%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "avg" + }, + { + "cacheTimeout": null, + "colorBackground": true, + "colorValue": false, + "colors": ["#967302", "#967302", "#967302"], + "description": "The number of cache request received", + "format": "short", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + "h": 3, + "w": 5, + "x": 4, + "y": 1 + }, + "id": 16, + "interval": null, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "150%", + "prefix": "", + "prefixFontSize": "150%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "sparkline": { + "fillColor": "#eab839", + "full": false, + "lineColor": "#e5ac0e", + "show": true + }, + "tableColumn": "", + "targets": [ + { + "expr": "registry_storage_cache_total{instance=\"$instance\",type=\"Request\"}", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ type }}", + "refId": "A" + } + ], + "thresholds": "", + "title": "Total Cache Requests", + "type": "singlestat", + "valueFontSize": "150%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "max" + }, + { + "cacheTimeout": null, + "colorBackground": true, + "colorValue": false, + "colors": ["#1f78c1", "#1f78c1", "#1f78c1"], + "description": "The number of cache request received (hits)", + "format": "short", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + "h": 3, + "w": 5, + "x": 9, + "y": 1 + }, + "id": 12, + "interval": null, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "150%", + "prefix": "", + "prefixFontSize": "150%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "sparkline": { + "fillColor": "#5195ce", + "full": false, + "lineColor": "#0a50a1", + "show": true + }, + "tableColumn": "", + "targets": [ + { + "expr": "registry_storage_cache_total{instance=\"$instance\",type=\"Hit\"}", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ type }}", + "refId": "A" + } + ], + "thresholds": "", + "title": "Total Hit Cache Requests", + "type": "singlestat", + "valueFontSize": "150%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "max" + }, + { + "cacheTimeout": null, + "colorBackground": true, + "colorValue": false, + "colors": ["#890f02", "#890f02", "#890f02"], + "description": "The number of cache request received (miss)", + "format": "short", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + "h": 3, + "w": 5, + "x": 14, + "y": 1 + }, + "id": 14, + "interval": null, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "150%", + "prefix": "", + "prefixFontSize": "150%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "sparkline": { + "fillColor": "#e24d42", + "full": false, + "lineColor": "#e24d42", + "show": true + }, + "tableColumn": "", + "targets": [ + { + "expr": "registry_storage_cache_total{instance=\"$instance\",type=\"Miss\"}", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ type }}", + "refId": "A" + } + ], + "thresholds": "", + "title": "Total Missed Cache Requests", + "type": "singlestat", + "valueFontSize": "150%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "avg" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": ["#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a"], + "description": "Process Resident Memory Usage", + "format": "decbytes", + "gauge": { + "maxValue": 1000, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + "h": 3, + "w": 5, + "x": 19, + "y": 1 + }, + "id": 24, + "interval": null, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "150%", + "prefix": "", + "prefixFontSize": "150%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": true, + "lineColor": "rgb(31, 120, 193)", + "show": true + }, + "tableColumn": "", + "targets": [ + { + "expr": "avg(process_resident_memory_bytes{instance=\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "", + "title": "Resident Memory Usage", + "type": "singlestat", + "valueFontSize": "150%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "avg" + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "description": "Average user and system CPU time spent in seconds.", + "fill": 1, + "gridPos": { + "h": 7, + "w": 8, + "x": 0, + "y": 4 + }, + "id": 30, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": true, + "min": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "avg(rate(process_cpu_seconds_total{instance=\"$instance\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "CPU Time", + "refId": "A" + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Average CPU Usage", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "description": "Virtual and Resident memory size in bytes, averages over 5 min interval", + "fill": 1, + "gridPos": { + "h": 7, + "w": 8, + "x": 8, + "y": 4 + }, + "id": 32, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": true, + "min": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "avg(rate(process_resident_memory_bytes{instance=\"$instance\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Resident Mem", + "refId": "A" + }, + { + "expr": "avg(rate(process_virtual_memory_bytes{instance=\"$instance\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Virtual Mem", + "refId": "B" + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Average Memory Usage", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "decimals": null, + "format": "decbytes", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "description": "Number of open file descriptors", + "fill": 1, + "gridPos": { + "h": 7, + "w": 8, + "x": 16, + "y": 4 + }, + "id": 34, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": true, + "min": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "process_open_fds{instance=\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Open FD", + "refId": "A" + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Open File Descriptors", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "decimals": null, + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 11 + }, + "id": 20, + "panels": [], + "title": "Registry Metrics", + "type": "row" + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "description": "The HTTP requests", + "fill": 1, + "gridPos": { + "h": 8, + "w": 8, + "x": 0, + "y": 12 + }, + "id": 26, + "legend": { + "alignAsTable": false, + "avg": false, + "current": true, + "max": false, + "min": false, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 3, + "links": [], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "ceil(rate(registry_http_requests_total{instance=\"$instance\"}[5m]))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ handler }}", + "refId": "A" + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "HTTP Requests", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": "0", + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "description": "The in-flight HTTP requests", + "fill": 1, + "gridPos": { + "h": 8, + "w": 8, + "x": 8, + "y": 12 + }, + "id": 18, + "legend": { + "alignAsTable": false, + "avg": false, + "current": true, + "max": false, + "min": false, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "registry_http_in_flight_requests{instance=\"$instance\"}", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ handler }}", + "refId": "A" + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "In-Flight HTTP Requests", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": "0", + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "description": "The rate of registry cache requests", + "fill": 1, + "gridPos": { + "h": 8, + "w": 8, + "x": 16, + "y": 12 + }, + "id": 28, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 3, + "links": [], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "ceil(rate(registry_storage_cache_total{instance=\"$instance\"}[5m]))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ type }}", + "refId": "A" + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Registry Cache Requests Rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": "0", + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "fill": 1, + "gridPos": { + "h": 7, + "w": 8, + "x": 0, + "y": 20 + }, + "id": 44, + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "increase(registry_storage_action_seconds_sum{instance=\"$instance\"}[2m]) * 1000", + "format": "time_series", + "instant": false, + "intervalFactor": 2, + "legendFormat": "{{ action }}", + "refId": "A", + "step": 10, + "target": "" + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "Registry Action Latency", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": "0", + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "cards": { + "cardPadding": null, + "cardRound": null + }, + "color": { + "cardColor": "#b4ff00", + "colorScale": "sqrt", + "colorScheme": "interpolateOranges", + "exponent": 0.5, + "mode": "spectrum" + }, + "dataFormat": "timeseries", + "gridPos": { + "h": 7, + "w": 8, + "x": 8, + "y": 20 + }, + "heatmap": {}, + "highlightCards": true, + "id": 36, + "legend": { + "show": false + }, + "links": [], + "targets": [ + { + "expr": "rate(registry_http_request_duration_seconds_bucket{handler=\"blob\"}[10m])", + "format": "heatmap", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "HTTP Request Latencies in seconds (blob)", + "tooltip": { + "show": true, + "showHistogram": false + }, + "type": "heatmap", + "xAxis": { + "show": true + }, + "xBucketNumber": null, + "xBucketSize": null, + "yAxis": { + "decimals": null, + "format": "short", + "logBase": 1, + "max": null, + "min": "0", + "show": true, + "splitFactor": null + }, + "yBucketBound": "auto", + "yBucketNumber": null, + "yBucketSize": null + }, + { + "cards": { + "cardPadding": null, + "cardRound": null + }, + "color": { + "cardColor": "#b4ff00", + "colorScale": "sqrt", + "colorScheme": "interpolateOranges", + "exponent": 0.5, + "mode": "spectrum" + }, + "dataFormat": "timeseries", + "gridPos": { + "h": 7, + "w": 8, + "x": 16, + "y": 20 + }, + "heatmap": {}, + "highlightCards": true, + "id": 38, + "legend": { + "show": false + }, + "links": [], + "targets": [ + { + "expr": "rate(registry_http_request_duration_seconds_bucket{handler=\"base\"}[10m])", + "format": "heatmap", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Upload HTTP Request Latencies in seconds (base)", + "tooltip": { + "show": true, + "showHistogram": false + }, + "type": "heatmap", + "xAxis": { + "show": true + }, + "xBucketNumber": null, + "xBucketSize": null, + "yAxis": { + "decimals": null, + "format": "short", + "logBase": 1, + "max": null, + "min": "0", + "show": true, + "splitFactor": null + }, + "yBucketBound": "auto", + "yBucketNumber": null, + "yBucketSize": null + }, + { + "cards": { + "cardPadding": null, + "cardRound": null + }, + "color": { + "cardColor": "#b4ff00", + "colorScale": "sqrt", + "colorScheme": "interpolateOranges", + "exponent": 0.5, + "mode": "spectrum" + }, + "dataFormat": "timeseries", + "gridPos": { + "h": 7, + "w": 8, + "x": 0, + "y": 27 + }, + "heatmap": {}, + "highlightCards": true, + "id": 2, + "legend": { + "show": false + }, + "links": [], + "targets": [ + { + "expr": "rate(registry_http_request_duration_seconds_bucket{handler=\"blob_upload\"}[10m])", + "format": "heatmap", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Upload HTTP Request Latencies in seconds (blob_upload)", + "tooltip": { + "show": true, + "showHistogram": false + }, + "type": "heatmap", + "xAxis": { + "show": true + }, + "xBucketNumber": null, + "xBucketSize": null, + "yAxis": { + "decimals": null, + "format": "short", + "logBase": 1, + "max": null, + "min": "0", + "show": true, + "splitFactor": null + }, + "yBucketBound": "auto", + "yBucketNumber": null, + "yBucketSize": null + }, + { + "cards": { + "cardPadding": null, + "cardRound": null + }, + "color": { + "cardColor": "#b4ff00", + "colorScale": "sqrt", + "colorScheme": "interpolateOranges", + "exponent": 0.5, + "mode": "spectrum" + }, + "dataFormat": "timeseries", + "gridPos": { + "h": 7, + "w": 8, + "x": 8, + "y": 27 + }, + "heatmap": {}, + "highlightCards": true, + "id": 42, + "legend": { + "show": false + }, + "links": [], + "targets": [ + { + "expr": "rate(registry_http_request_duration_seconds_bucket{handler=\"manifest\"}[10m])", + "format": "heatmap", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Manifest HTTP Request Latencies in seconds (manifest)", + "tooltip": { + "show": true, + "showHistogram": false + }, + "type": "heatmap", + "xAxis": { + "show": true + }, + "xBucketNumber": null, + "xBucketSize": null, + "yAxis": { + "decimals": null, + "format": "short", + "logBase": 1, + "max": null, + "min": "0", + "show": true, + "splitFactor": null + }, + "yBucketBound": "auto", + "yBucketNumber": null, + "yBucketSize": null + }, + { + "cards": { + "cardPadding": null, + "cardRound": null + }, + "color": { + "cardColor": "#b4ff00", + "colorScale": "sqrt", + "colorScheme": "interpolateOranges", + "exponent": 0.5, + "mode": "spectrum" + }, + "dataFormat": "timeseries", + "gridPos": { + "h": 7, + "w": 8, + "x": 16, + "y": 27 + }, + "heatmap": {}, + "highlightCards": true, + "id": 40, + "legend": { + "show": false + }, + "links": [], + "targets": [ + { + "expr": "rate(registry_http_request_duration_seconds_bucket{handler=\"catalog\"}[10m])", + "format": "heatmap", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Catalog HTTP Request Latencies in seconds (catalog)", + "tooltip": { + "show": true, + "showHistogram": false + }, + "type": "heatmap", + "xAxis": { + "show": true + }, + "xBucketNumber": null, + "xBucketSize": null, + "yAxis": { + "decimals": null, + "format": "short", + "logBase": 1, + "max": null, + "min": "0", + "show": true, + "splitFactor": null + }, + "yBucketBound": "auto", + "yBucketNumber": null, + "yBucketSize": null + } + ], + "refresh": "10s", + "schemaVersion": 16, + "style": "dark", + "tags": ["docker", "registry", "internals"], + "templating": { + "list": [ + { + "allValue": null, + "current": {}, + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "namespace", + "options": [], + "query": "query_result(registry_http_in_flight_requests)", + "refresh": 2, + "regex": "/.*kubernetes_namespace=\"([^\"]+).*/", + "sort": 0, + "tagValuesQuery": "", + "tags": [], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": {}, + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "app", + "options": [], + "query": "query_result(registry_http_in_flight_requests{kubernetes_namespace=\"$namespace\"})", + "refresh": 2, + "regex": "/.*app=\"([^\"]+)/", + "sort": 0, + "tagValuesQuery": "", + "tags": [], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": {}, + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "instance", + "options": [], + "query": "query_result(up{app=\"$app\"})", + "refresh": 1, + "regex": "/.*instance=\"([^\"]+).*/", + "sort": 0, + "tagValuesQuery": "", + "tags": [], + "tagsQuery": "", + "type": "query", + "useTags": false + } + ] + }, + "time": { + "from": "now-6h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": ["5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d"] + }, + "timezone": "", + "title": "Docker Registry", + "uid": "CoBSgj8iz", + "version": 19 +} diff --git a/rootfs/usr/share/grafana/dashboards/drycc_storage.json b/rootfs/usr/share/grafana/dashboards/drycc_storage.json new file mode 100644 index 0000000..4926192 --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/drycc_storage.json @@ -0,0 +1,3258 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "target": { + "limit": 100, + "matchAny": false, + "tags": [], + "type": "dashboard" + }, + "type": "dashboard" + } + ] + }, + "description": "MinIO Grafana Dashboard - https://min.io/", + "editable": false, + "fiscalYearStartMonth": 0, + "gnetId": 13502, + "graphTooltip": 0, + "id": null, + "links": [ + { + "icon": "external link", + "includeVars": true, + "keepTime": true, + "tags": ["minio"], + "type": "dashboards" + } + ], + "liveNow": false, + "panels": [ + { + "description": "", + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "percentage", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "dtdurations", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 6, + "w": 3, + "x": 0, + "y": 0 + }, + "id": 1, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "exemplar": true, + "expr": "time() - max(minio_node_process_starttime_seconds{job=~\"$scrape_jobs\"})", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{instance}}", + "metric": "process_start_time_seconds", + "refId": "A", + "step": 60 + } + ], + "title": "Uptime", + "type": "stat" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "bytes", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 3, + "x": 3, + "y": 0 + }, + "id": 65, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "exemplar": true, + "expr": "sum by (instance) (minio_s3_traffic_received_bytes{job=~\"$scrape_jobs\"})", + "format": "table", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{instance}}", + "metric": "process_start_time_seconds", + "refId": "A", + "step": 60 + } + ], + "title": "Total S3 Ingress", + "type": "stat" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + } + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "unit": "bytes", + "unitScale": true + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Free" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "green", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Used" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "orange", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 6, + "w": 4, + "x": 6, + "y": 0 + }, + "id": 50, + "interval": "1m", + "links": [], + "maxDataPoints": 100, + "options": { + "displayLabels": [], + "legend": { + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "values": ["percent"] + }, + "pieType": "donut", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "8.2.1", + "targets": [ + { + "exemplar": true, + "expr": "topk(1, sum(minio_cluster_capacity_usable_total_bytes{job=~\"$scrape_jobs\"}) by (instance)) - topk(1, sum(minio_cluster_capacity_usable_free_bytes{job=~\"$scrape_jobs\"}) by (instance))", + "format": "time_series", + "instant": false, + "interval": "1m", + "intervalFactor": 1, + "legendFormat": "Used", + "refId": "A", + "step": 300 + }, + { + "exemplar": true, + "expr": "topk(1, sum(minio_cluster_capacity_usable_free_bytes{job=~\"$scrape_jobs\"}) by (instance)) ", + "hide": false, + "interval": "1m", + "legendFormat": "Free", + "refId": "B" + } + ], + "title": "Capacity", + "type": "piechart" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": true, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes", + "unitScale": true + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Objects" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "blue", + "mode": "fixed" + } + } + ] + }, + { + "__systemRef": "hideSeriesFrom", + "matcher": { + "id": "byNames", + "options": { + "mode": "exclude", + "names": ["Usage"], + "prefix": "All except:", + "readOnly": true + } + }, + "properties": [ + { + "id": "custom.hideFrom", + "value": { + "legend": false, + "tooltip": false, + "viz": true + } + } + ] + } + ] + }, + "gridPos": { + "h": 6, + "w": 6, + "x": 10, + "y": 0 + }, + "id": 68, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "8.2.1", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "max(minio_cluster_usage_total_bytes{job=~\"$scrape_jobs\"})", + "interval": "", + "legendFormat": "Usage", + "range": true, + "refId": "A" + } + ], + "title": "Data Usage Growth", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "continuous-GrYlRd" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 6, + "w": 5, + "x": 16, + "y": 0 + }, + "id": 52, + "links": [], + "options": { + "displayMode": "lcd", + "maxVizHeight": 300, + "minVizHeight": 16, + "minVizWidth": 8, + "namePlacement": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showUnfilled": true, + "sizing": "auto", + "valueMode": "color" + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "disableTextWrap": false, + "editorMode": "code", + "exemplar": true, + "expr": "minio_cluster_objects_size_distribution{job=~\"$scrape_jobs\"}", + "format": "time_series", + "fullMetaSearch": false, + "includeNullMetadata": true, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{range}}", + "refId": "A", + "step": 300, + "useBackend": false + } + ], + "title": "Object Size Distribution", + "type": "bargauge" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 25, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 3, + "x": 21, + "y": 0 + }, + "id": 61, + "links": [], + "maxDataPoints": 100, + "options": { + "legend": { + "calcs": [], + "displayMode": "hidden", + "placement": "right", + "showLegend": false + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "minio_node_file_descriptor_open_total{job=~\"$scrape_jobs\"}", + "format": "time_series", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{server}}", + "metric": "process_start_time_seconds", + "refId": "A", + "step": 60 + } + ], + "title": "Open FDs ", + "type": "timeseries" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "bytes", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 3, + "x": 3, + "y": 3 + }, + "id": 64, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "exemplar": true, + "expr": "sum by (instance) (minio_s3_traffic_sent_bytes{job=~\"$scrape_jobs\"})", + "format": "table", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "metric": "process_start_time_seconds", + "refId": "A", + "step": 60 + } + ], + "title": "Total S3 Egress", + "type": "stat" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 25, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 3, + "x": 21, + "y": 3 + }, + "id": 62, + "links": [], + "maxDataPoints": 100, + "options": { + "legend": { + "calcs": [], + "displayMode": "hidden", + "placement": "right", + "showLegend": false + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "minio_node_go_routine_total{job=~\"$scrape_jobs\"}", + "format": "time_series", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{server}}", + "metric": "process_start_time_seconds", + "refId": "A", + "step": 60 + } + ], + "title": "Goroutines", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "bool_on_off", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 2, + "w": 3, + "x": 0, + "y": 6 + }, + "id": 94, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "disableTextWrap": false, + "editorMode": "code", + "exemplar": true, + "expr": "minio_cluster_health_status{job=~\"$scrape_jobs\"}", + "fullMetaSearch": false, + "includeNullMetadata": true, + "interval": "", + "legendFormat": "Pool: {{pool}} Set: {{set}}", + "range": true, + "refId": "A", + "useBackend": false + } + ], + "title": "Cluster Health Status", + "transformations": [], + "type": "stat" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "mappings": [], + "thresholds": { + "mode": "percentage", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 3, + "x": 3, + "y": 6 + }, + "id": 78, + "links": [], + "maxDataPoints": 100, + "options": { + "minVizHeight": 75, + "minVizWidth": 75, + "orientation": "auto", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showThresholdLabels": false, + "showThresholdMarkers": true, + "sizing": "auto" + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "editorMode": "code", + "exemplar": false, + "expr": "max(minio_cluster_drive_online_total{job=~\"$scrape_jobs\"})", + "format": "time_series", + "hide": false, + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": ".", + "metric": "process_start_time_seconds", + "range": false, + "refId": "A", + "step": 60 + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "max(minio_cluster_drive_offline_total{job=~\"$scrape_jobs\"})", + "format": "time_series", + "hide": false, + "instant": true, + "legendFormat": ".", + "range": false, + "refId": "B" + } + ], + "title": "Total Online/Offline Drives", + "type": "gauge" + }, + { + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "dark-yellow", + "value": 75000000 + }, + { + "color": "dark-red", + "value": 100000000 + } + ] + }, + "unit": "short", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 3, + "x": 6, + "y": 6 + }, + "id": 66, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "max(minio_cluster_bucket_total{job=~\"$scrape_jobs\"})", + "format": "time_series", + "instant": false, + "interval": "1m", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Number of Buckets", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 25, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "binBps", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 6, + "w": 7, + "x": 9, + "y": 6 + }, + "id": 63, + "options": { + "legend": { + "calcs": [], + "displayMode": "hidden", + "placement": "right", + "showLegend": false + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "exemplar": true, + "expr": "sum by (server) (rate(minio_s3_traffic_received_bytes{job=~\"$scrape_jobs\"}[$__rate_interval]))", + "interval": "1m", + "intervalFactor": 2, + "legendFormat": "Data Received [{{server}}]", + "refId": "A" + } + ], + "title": "S3 API Ingress Rate ", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 25, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "binBps", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 6, + "w": 8, + "x": 16, + "y": 6 + }, + "id": 70, + "options": { + "legend": { + "calcs": [], + "displayMode": "hidden", + "placement": "right", + "showLegend": false + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "exemplar": true, + "expr": "sum by (server) (rate(minio_s3_traffic_sent_bytes{job=~\"$scrape_jobs\"}[$__rate_interval]))", + "interval": "1m", + "intervalFactor": 2, + "legendFormat": "Data Sent [{{server}}]", + "refId": "A" + } + ], + "title": "S3 API Egress Rate ", + "type": "timeseries" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 2, + "w": 3, + "x": 0, + "y": 8 + }, + "id": 53, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "exemplar": true, + "expr": "max(minio_cluster_nodes_online_total{job=~\"$scrape_jobs\"})", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "metric": "process_start_time_seconds", + "refId": "A", + "step": 60 + } + ], + "title": "Total Online Servers", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "dark-yellow", + "value": 75000000 + }, + { + "color": "dark-red", + "value": 100000000 + } + ] + }, + "unit": "short", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 3, + "x": 6, + "y": 9 + }, + "id": 44, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "max(minio_cluster_usage_object_total{job=~\"$scrape_jobs\"})", + "format": "time_series", + "instant": false, + "interval": "1m", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Number of Objects", + "type": "stat" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "ns", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 2, + "w": 3, + "x": 0, + "y": 10 + }, + "id": 80, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "exemplar": true, + "expr": "max(minio_heal_time_last_activity_nano_seconds{job=~\"$scrape_jobs\"})", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "metric": "process_start_time_seconds", + "refId": "A", + "step": 60 + } + ], + "title": "Time Since Last Heal", + "type": "stat" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "ns", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 2, + "w": 3, + "x": 3, + "y": 10 + }, + "id": 81, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "exemplar": true, + "expr": "max(minio_usage_last_activity_nano_seconds{job=~\"$scrape_jobs\"})", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "metric": "process_start_time_seconds", + "refId": "A", + "step": 60 + } + ], + "title": "Time Since Last Scan", + "type": "stat" + }, + { + "aliasColors": { + "S3 Errors": "light-red", + "S3 Requests": "light-green" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 6, + "w": 9, + "x": 0, + "y": 12 + }, + "hiddenSeries": false, + "id": 60, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "sum by (server,api) (increase(minio_s3_requests_total{job=~\"$scrape_jobs\"}[$__rate_interval]))", + "interval": "1m", + "intervalFactor": 2, + "legendFormat": "{{server,api}}", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "S3 API Request Rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:331", + "format": "none", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:332", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": { + "S3 Errors": "light-red", + "S3 Requests": "light-green" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 6, + "w": 7, + "x": 9, + "y": 12 + }, + "hiddenSeries": false, + "id": 88, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "sum by (server,api) (increase(minio_s3_requests_4xx_errors_total{job=~\"$scrape_jobs\"}[$__rate_interval]))", + "interval": "1m", + "intervalFactor": 2, + "legendFormat": "{{server,api}}", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "S3 API Request Error Rate (4xx)", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:331", + "format": "none", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:332", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": { + "S3 Errors": "light-red", + "S3 Requests": "light-green" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 6, + "w": 8, + "x": 16, + "y": 12 + }, + "hiddenSeries": false, + "id": 86, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "sum by (server,api) (increase(minio_s3_requests_5xx_errors_total{job=~\"$scrape_jobs\"}[$__rate_interval]))", + "interval": "1m", + "intervalFactor": 2, + "legendFormat": "{{server,api}}", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "S3 API Request Error Rate (5xx)", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:331", + "format": "none", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:332", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "fixed" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 18 + }, + "id": 99, + "options": { + "displayMode": "lcd", + "maxVizHeight": 300, + "minVizHeight": 16, + "minVizWidth": 8, + "namePlacement": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showUnfilled": true, + "sizing": "auto", + "valueMode": "color" + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "disableTextWrap": false, + "editorMode": "builder", + "expr": "minio_cluster_health_erasure_set_online_drives{job=~\"$scrape_jobs\"}", + "fullMetaSearch": false, + "hide": false, + "includeNullMetadata": true, + "instant": false, + "legendFormat": "Pool {{pool}} / Set {{set}} - Online Drives", + "range": true, + "refId": "A", + "useBackend": false + }, + { + "disableTextWrap": false, + "editorMode": "builder", + "expr": "minio_cluster_health_erasure_set_read_quorum{job=~\"$scrape_jobs\"}", + "fullMetaSearch": false, + "hide": false, + "includeNullMetadata": true, + "instant": false, + "legendFormat": "Pool {{pool}} / Set {{set}} - Read Quorum", + "range": true, + "refId": "B", + "useBackend": false + }, + { + "disableTextWrap": false, + "editorMode": "builder", + "expr": "minio_cluster_health_erasure_set_write_quorum{job=~\"$scrape_jobs\"}", + "fullMetaSearch": false, + "hide": false, + "includeNullMetadata": true, + "instant": false, + "legendFormat": "Pool {{pool}} / Set {{set}} - Write Quorum", + "range": true, + "refId": "C", + "useBackend": false + }, + { + "disableTextWrap": false, + "editorMode": "builder", + "expr": "minio_cluster_health_erasure_set_healing_drives{job=~\"$scrape_jobs\"}", + "fullMetaSearch": false, + "hide": false, + "includeNullMetadata": true, + "instant": false, + "legendFormat": "Pool {{pool}} / Set {{set}} - Healing Drives", + "range": true, + "refId": "D", + "useBackend": false + }, + { + "disableTextWrap": false, + "editorMode": "builder", + "expr": "minio_cluster_health_erasure_set_status{job=~\"$scrape_jobs\"}", + "fullMetaSearch": false, + "hide": false, + "includeNullMetadata": true, + "instant": false, + "legendFormat": "Pool {{pool}} / Set {{set}} - Status", + "range": true, + "refId": "E", + "useBackend": false + } + ], + "title": "Health Breakdown", + "type": "bargauge" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "continuous-GrYlRd" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 18 + }, + "id": 76, + "options": { + "displayMode": "lcd", + "maxVizHeight": 300, + "minVizHeight": 16, + "minVizWidth": 8, + "namePlacement": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showUnfilled": true, + "sizing": "auto", + "valueMode": "color" + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "editorMode": "code", + "exemplar": false, + "expr": "minio_node_process_resident_memory_bytes{job=~\"$scrape_jobs\"}", + "format": "time_series", + "instant": false, + "interval": "", + "legendFormat": "{{server}}", + "range": true, + "refId": "A" + } + ], + "title": "Memory Usage ", + "type": "bargauge" + }, + { + "aliasColors": {}, + "bars": true, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 26 + }, + "hiddenSeries": false, + "id": 73, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "rate(minio_node_io_rchar_bytes{job=~\"$scrape_jobs\"}[$__rate_interval])", + "format": "time_series", + "instant": false, + "interval": "", + "legendFormat": "Node RChar [{{server}}]", + "refId": "A" + }, + { + "exemplar": true, + "expr": "rate(minio_node_io_wchar_bytes{job=~\"$scrape_jobs\"}[$__rate_interval])", + "interval": "", + "legendFormat": "Node WChar [{{server}}]", + "refId": "B" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Read, Write I/O", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:381", + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:382", + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "fieldConfig": { + "defaults": { + "mappings": [], + "thresholds": { + "mode": "percentage", + "steps": [ + { + "color": "green" + }, + { + "color": "orange", + "value": 70 + }, + { + "color": "red", + "value": 85 + } + ] + }, + "unit": "s", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 26 + }, + "id": 77, + "options": { + "minVizHeight": 75, + "minVizWidth": 75, + "orientation": "auto", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showThresholdLabels": false, + "showThresholdMarkers": true, + "sizing": "auto" + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "rate(minio_node_process_cpu_total_seconds{job=~\"$scrape_jobs\"}[$__rate_interval])", + "interval": "", + "legendFormat": "{{server}}", + "range": true, + "refId": "A" + } + ], + "title": "CPU Usage", + "type": "gauge" + }, + { + "description": "Total number of bytes received and sent on MinIO cluster", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 25, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 33 + }, + "id": 17, + "links": [], + "options": { + "legend": { + "calcs": [], + "displayMode": "hidden", + "placement": "right", + "showLegend": false + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "rate(minio_inter_node_traffic_sent_bytes{job=~\"$scrape_jobs\"}[$__rate_interval])", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "Internode Bytes Received [{{server}}]", + "metric": "minio_http_requests_duration_seconds_count", + "range": true, + "refId": "A", + "step": 4 + }, + { + "exemplar": true, + "expr": "rate(minio_inter_node_traffic_received_bytes{job=~\"$scrape_jobs\"}[$__rate_interval])", + "interval": "", + "legendFormat": "Internode Bytes Sent [{{server}}]", + "refId": "B" + } + ], + "title": "Internode Traffic", + "type": "timeseries" + }, + { + "aliasColors": { + "available 10.13.1.25:9000": "green", + "used 10.13.1.25:9000": "blue" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "description": "", + "fieldConfig": { + "defaults": { + "links": [], + "unit": "bytes", + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 33 + }, + "hiddenSeries": false, + "id": 8, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "minio_node_file_descriptor_open_total{job=~\"$scrape_jobs\"}", + "interval": "", + "legendFormat": "Open FDs [{{server}}]", + "refId": "B" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "File Descriptors", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:212", + "format": "bytes", + "logBase": 1, + "min": "0", + "show": true + }, + { + "$$hashKey": "object:213", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": { + "Offline 10.13.1.25:9000": "dark-red", + "Total 10.13.1.25:9000": "blue" + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "description": "Number of online drives per MinIO Server", + "fieldConfig": { + "defaults": { + "links": [], + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 40 + }, + "hiddenSeries": false, + "id": 11, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": false, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "rate(minio_node_syscall_read_total{job=~\"$scrape_jobs\"}[$__rate_interval])", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "Read Syscalls [{{server}}]", + "metric": "process_start_time_seconds", + "refId": "A", + "step": 60 + }, + { + "exemplar": true, + "expr": "rate(minio_node_syscall_write_total{job=~\"$scrape_jobs\"}[$__rate_interval])", + "interval": "", + "legendFormat": "Write Syscalls [{{server}}]", + "refId": "B" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Syscalls", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:185", + "decimals": 0, + "format": "short", + "logBase": 1, + "min": "0", + "show": true + }, + { + "$$hashKey": "object:186", + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 40 + }, + "hiddenSeries": false, + "id": 95, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "rate(minio_node_scanner_objects_scanned{job=~\"$scrape_jobs\"}[$__rate_interval])", + "interval": "1m", + "legendFormat": "[{{server}}]", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Scanned Objects", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:212", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + }, + { + "$$hashKey": "object:213", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 47 + }, + "hiddenSeries": false, + "id": 75, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "rate(minio_node_scanner_versions_scanned{job=~\"$scrape_jobs\"}[$__rate_interval])", + "interval": "1m", + "legendFormat": "[{{server}}]", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Scanned Versions", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:212", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + }, + { + "$$hashKey": "object:213", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 47 + }, + "hiddenSeries": false, + "id": 96, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "rate(minio_node_scanner_directories_scanned{job=~\"$scrape_jobs\"}[$__rate_interval])", + "interval": "1m", + "legendFormat": "[{{server}}]", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Scanned Directories", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:212", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + }, + { + "$$hashKey": "object:213", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "percentage", + "steps": [ + { + "color": "green" + } + ] + }, + "unit": "dtdurations", + "unitScale": true + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 4, + "x": 0, + "y": 54 + }, + "id": 89, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.3.1", + "targets": [ + { + "exemplar": true, + "expr": "minio_cluster_kms_uptime{job=~\"$scrape_jobs\"}", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{instance}}", + "metric": "minio_cluster_kms_uptime", + "refId": "A", + "step": 60 + } + ], + "title": "KMS Uptime", + "type": "stat" + }, + { + "aliasColors": { + "S3 Errors": "light-red", + "S3 Requests": "light-green" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 4, + "w": 4, + "x": 4, + "y": 54 + }, + "hiddenSeries": false, + "id": 91, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "sum by (server) (increase(minio_cluster_kms_request_error{job=~\"$scrape_jobs\"}[$__rate_interval]))", + "interval": "1m", + "intervalFactor": 2, + "legendFormat": "{{server}}", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "KMS Request 4xx Error Rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:331", + "format": "none", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:332", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unit": "bool_on_off", + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 4, + "w": 4, + "x": 8, + "y": 54 + }, + "hiddenSeries": false, + "id": 90, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "sum by (server) (minio_cluster_kms_online{job=~\"$scrape_jobs\"})", + "interval": "1m", + "legendFormat": "{{server}}", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "KMS Online(1)/Offline(0)", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:212", + "format": "bool_on_off", + "logBase": 1, + "min": "0", + "show": true + }, + { + "$$hashKey": "object:213", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 54 + }, + "hiddenSeries": false, + "id": 98, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "rate(minio_node_scanner_bucket_scans_finished{job=~\"$scrape_jobs\"}[$__rate_interval])", + "interval": "1m", + "legendFormat": "[{{server}}]", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Bucket Scans Finished", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:212", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + }, + { + "$$hashKey": "object:213", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": { + "S3 Errors": "light-red", + "S3 Requests": "light-green" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 5, + "w": 6, + "x": 0, + "y": 58 + }, + "hiddenSeries": false, + "id": 92, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "sum by (server) (increase(minio_cluster_kms_request_failure{job=~\"$scrape_jobs\"}[$__rate_interval]))", + "interval": "1m", + "intervalFactor": 2, + "legendFormat": "{{server}}", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "KMS Request 5xx Error Rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:331", + "format": "none", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:332", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": { + "S3 Errors": "light-red", + "S3 Requests": "light-green" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 5, + "w": 6, + "x": 6, + "y": 58 + }, + "hiddenSeries": false, + "id": 93, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "sum by (server) (rate(minio_cluster_kms_request_success{job=~\"$scrape_jobs\"}[$__rate_interval]))", + "interval": "1m", + "intervalFactor": 2, + "legendFormat": "KMS Request Success [{{server}}]", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "KMS Request Success Rate ", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:331", + "format": "none", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:332", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "fieldConfig": { + "defaults": { + "unitScale": true + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 61 + }, + "hiddenSeries": false, + "id": 97, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "10.3.1", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "exemplar": true, + "expr": "rate(minio_node_scanner_bucket_scans_started{job=~\"$scrape_jobs\"}[$__rate_interval])", + "interval": "1m", + "legendFormat": "[{{server}}]", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Bucket Scans Started", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:212", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + }, + { + "$$hashKey": "object:213", + "format": "none", + "logBase": 1, + "min": "0", + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "refresh": "", + "schemaVersion": 39, + "tags": ["minio"], + "templating": { + "list": [ + { + "current": {}, + "definition": "label_values(job)", + "hide": 0, + "includeAll": true, + "multi": true, + "name": "scrape_jobs", + "options": [], + "query": { + "query": "label_values(job)", + "refId": "StandardVariableQuery" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "type": "query" + } + ] + }, + "time": { + "from": "now-6h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": ["5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d"] + }, + "timezone": "", + "title": "MinIO Dashboard", + "uid": "TgmJnqnnk", + "version": 54, + "weekStart": "" +} diff --git a/rootfs/usr/share/grafana/dashboards/drycc_valkey.json b/rootfs/usr/share/grafana/dashboards/drycc_valkey.json new file mode 100644 index 0000000..07b0982 --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/drycc_valkey.json @@ -0,0 +1,1364 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": "-- Grafana --", + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "description": "Redis Dashboard for Prometheus Redis Exporter 1.x", + "editable": false, + "gnetId": 763, + "graphTooltip": 1, + "id": null, + "iteration": 1602758020790, + "links": [], + "panels": [ + { + "collapsed": false, + "datasource": null, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 24, + "panels": [], + "title": "Performance", + "type": "row" + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "description": "Average taken across instances", + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 8, + "x": 0, + "y": 1 + }, + "hiddenSeries": false, + "id": 18, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "hideEmpty": false, + "hideZero": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "connected", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "avg(irate(redis_commands_total{instance=~\"$instance\"} [$__rate_interval])) by (cmd)", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{cmd}}", + "metric": "redis_command_calls_total", + "refId": "A", + "step": 240 + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Commands per second", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "description": "Average taken across instances", + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 8, + "x": 8, + "y": 1 + }, + "hiddenSeries": false, + "id": 20, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "hideEmpty": false, + "hideZero": true, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "connected", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "avg(irate(redis_commands_duration_seconds_total{instance=~\"$instance\"}[$__rate_interval])) by (cmd)\n /\navg(irate(redis_commands_total{instance=~\"$instance\"}[$__rate_interval])) by (cmd)\n", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{ cmd }}", + "metric": "redis_command_calls_total", + "refId": "A", + "step": 240 + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Command latency per second", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": { + "Hit ratio": "blue" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "decimals": 2, + "description": "Hit rate shows the percentage of key space lookups that hit a key.", + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 8, + "x": 16, + "y": 1 + }, + "hiddenSeries": false, + "id": 1, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "connected", + "options": { + "dataLinks": [] + }, + "percentage": true, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + { + "alias": "/Target/", + "color": "#56A64B", + "dashes": true, + "fill": 0, + "hideTooltip": true, + "linewidth": 1 + } + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "avg(irate(redis_keyspace_hits_total{instance=~\"$instance\"}[$__rate_interval]) / (irate(redis_keyspace_misses_total{instance=~\"$instance\"}[$__rate_interval]) + irate(redis_keyspace_hits_total{instance=~\"$instance\"}[$__rate_interval]))) by (instance)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "metric": "", + "refId": "A", + "step": 240, + "target": "" + }, + { + "expr": "1", + "interval": "", + "legendFormat": "Target hit ratio for cache", + "refId": "B" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Hit ratio per instance", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "percentunit", + "label": "", + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "collapsed": false, + "datasource": null, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 8 + }, + "id": 22, + "panels": [], + "title": "Memory", + "type": "row" + }, + { + "aliasColors": { + "max": "#BF1B00" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "description": "Total taken across instances", + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 8, + "x": 0, + "y": 9 + }, + "hiddenSeries": false, + "id": 7, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "hideEmpty": false, + "hideZero": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null as zero", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + { + "alias": "/max/", + "color": "#E02F44", + "dashes": true, + "fill": 0, + "linewidth": 1 + } + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(redis_memory_used_bytes{instance=~\"$instance\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "Used Memory", + "metric": "", + "refId": "A", + "step": 240, + "target": "" + }, + { + "expr": "sum(redis_memory_max_bytes{instance=~\"$instance\"})", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 2, + "legendFormat": "Configured max memory", + "refId": "B", + "step": 240 + }, + { + "expr": "sum(redis_memory_used_rss_bytes{instance=~\"$instance\"})", + "interval": "", + "legendFormat": "Used RSS memory", + "refId": "C" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Total Memory Usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": { + "Recommend restart redis": "red" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 8, + "x": 8, + "y": 9 + }, + "hiddenSeries": false, + "id": 10, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "connected", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + { + "alias": "/restart/", + "color": "#E02F44", + "dashes": true, + "fill": 0, + "linewidth": 1 + } + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "redis_memory_fragmentation_ratio{instance=~\"$instance\"}", + "hide": false, + "interval": "", + "legendFormat": "{{instance}}", + "refId": "C" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Memory fragmentation ratio per instance", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": { + "Evictions": "red", + "evicts": "#890F02", + "memcached_items_evicted_total{instance=\"172.17.0.1:9150\",job=\"prometheus\"}": "#890F02", + "reclaims": "#3F6833", + "{container=\"redis-exporter\", instance=\"redis-86cb5d76d7-fcdln:redis-exporter:redis-metrics\", job=\"default/redis\", namespace=\"default\", pod=\"redis-86cb5d76d7-fcdln\"}": "red", + "{instance=\"redis-86cb5d76d7-fcdln:redis-exporter:redis-metrics\"}": "red" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 8, + "x": 16, + "y": 9 + }, + "hiddenSeries": false, + "id": 8, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "connected", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + { + "alias": "reclaims", + "yaxis": 2 + } + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "irate(redis_evicted_keys_total{instance=~\"$instance\"}[$__rate_interval])", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "B", + "step": 240 + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Key evictions per second per instance", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "collapsed": false, + "datasource": null, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 16 + }, + "id": 26, + "panels": [], + "title": "Basic activity", + "type": "row" + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "description": "Sum taken across instances", + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 8, + "x": 0, + "y": 17 + }, + "hiddenSeries": false, + "id": 16, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(redis_connected_clients{instance=~\"$instance\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "Connected", + "refId": "A" + }, + { + "expr": "sum(redis_blocked_clients{instance=~\"$instance\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "Blocked", + "refId": "B" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Connected/Blocked Clients", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": { + "db1": "yellow" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "description": "Sum taken across instances", + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 8, + "x": 8, + "y": 17 + }, + "hiddenSeries": false, + "id": 5, + "isNew": true, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": false, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum (redis_db_keys{instance=~\"$instance\"}) by (db)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{ db }}", + "refId": "A", + "step": 240, + "target": "" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Total Items per DB", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 1, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "none", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "description": "Sum taken across instances", + "editable": false, + "error": false, + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 8, + "x": 16, + "y": 17 + }, + "hiddenSeries": false, + "id": 13, + "isNew": true, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "connected", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum (redis_db_keys{instance=~\"$instance\"}) - sum (redis_db_keys_expiring{instance=~\"$instance\"})", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 2, + "legendFormat": "Not expiring", + "refId": "A", + "step": 240, + "target": "" + }, + { + "expr": "sum(redis_db_keys_expiring{instance=~\"$instance\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "Expiring", + "metric": "", + "refId": "B", + "step": 240 + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Expiring vs Not-Expiring Keys", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "description": "This metric will only be non-zero if the instance is a master", + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 8, + "x": 0, + "y": 24 + }, + "hiddenSeries": false, + "id": 28, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(redis_connected_slaves{instance=~\"$instance\"}) by (instance)", + "interval": "", + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Connected slaves by instance", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "description": "This metric is only exported if the instance is a slave.", + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 8, + "x": 8, + "y": 24 + }, + "hiddenSeries": false, + "id": 30, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "redis_master_last_io_seconds_ago{instance=~\"$instance\"}", + "interval": "", + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "Time since last master connection", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + } + ], + "refresh": false, + "schemaVersion": 25, + "style": "dark", + "tags": ["prometheus", "redis"], + "templating": { + "list": [ + { + "allValue": null, + "current": { + "selected": true, + "tags": [], + "text": "redisdb-7d6b98cd98-kjt5x + redisdb-7d6b98cd98-pbkg6", + "value": ["redisdb-7d6b98cd98-kjt5x", "redisdb-7d6b98cd98-pbkg6"] + }, + "datasource": "${datasource}", + "definition": "label_values(redis_up, instance)", + "hide": 0, + "includeAll": false, + "label": null, + "multi": true, + "name": "instance", + "options": [], + "query": "label_values(redis_up, instance)", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "current": { + "selected": false, + "text": "prometheus", + "value": "prometheus" + }, + "hide": 0, + "includeAll": false, + "label": "Data Source", + "multi": false, + "name": "datasource", + "options": [], + "query": "prometheus", + "queryValue": "", + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "type": "datasource" + } + ] + }, + "time": { + "from": "now-12h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": ["5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d"] + }, + "timezone": "browser", + "title": "Redis Dashboard for Prometheus Redis Exporter 1.x", + "uid": "bRd48yKMdd", + "version": 5 +} diff --git a/rootfs/usr/share/grafana/dashboards/drycc_victoriametrics_cluster.json b/rootfs/usr/share/grafana/dashboards/drycc_victoriametrics_cluster.json new file mode 100644 index 0000000..0f35028 --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/drycc_victoriametrics_cluster.json @@ -0,0 +1,9920 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "target": { + "limit": 100, + "matchAny": false, + "tags": [], + "type": "dashboard" + }, + "type": "dashboard" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "enable": true, + "expr": "sum(ALERTS{job=~\"$job\", instance=~\"$instance\", alertgroup=\"vmcluster\",alertstate=\"firing\",show_at=\"dashboard\"}) by(alertname)", + "hide": false, + "iconColor": "red", + "name": "alerts", + "titleFormat": "{{alertname}}" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "enable": true, + "expr": "sum(vm_app_version{job=~\"$job\", instance=~\"$instance\"}) by(version) unless (sum(vm_app_version{job=~\"$job\", instance=~\"$instance\"} offset $__interval) by(version))", + "hide": true, + "iconColor": "dark-blue", + "name": "version change", + "textFormat": "{{version}}", + "titleFormat": "Version change" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "enable": true, + "expr": "sum(changes(vm_app_start_timestamp{job=~\"$job\", instance=~\"$instance\"}[$__interval])) by(job)", + "hide": false, + "iconColor": "dark-yellow", + "name": "restarts", + "textFormat": "{{job}} restarted" + } + ] + }, + "description": "Overview for cluster VictoriaMetrics v1.102.0 or higher", + "editable": false, + "fiscalYearStartMonth": 0, + "graphTooltip": 1, + "id": 1, + "links": [ + { + "icon": "doc", + "tags": [], + "targetBlank": true, + "title": "Cluster Wiki", + "type": "link", + "url": "https://docs.victoriametrics.com/cluster-victoriametrics" + }, + { + "icon": "external link", + "tags": [], + "targetBlank": true, + "title": "Found a bug?", + "type": "link", + "url": "https://github.com/VictoriaMetrics/VictoriaMetrics/issues" + }, + { + "icon": "external link", + "tags": [], + "targetBlank": true, + "title": "New releases", + "type": "link", + "url": "https://github.com/VictoriaMetrics/VictoriaMetrics/releases" + } + ], + "panels": [ + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 137, + "panels": [], + "title": "Stats", + "type": "row" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "How many [data samples](https://docs.victoriametrics.com/keyconcepts/#raw-samples) are in storage", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 0, + "y": 1 + }, + "id": 131, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": true, + "expr": "sum(vm_rows{job=~\"$job_storage\", type!~\"indexdb.*\"})", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Total datapoints", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the datapoints ingestion rate, including replication factor.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 6, + "y": 1 + }, + "id": 124, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "sum(rate(vm_vminsert_metrics_read_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])) ", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "range": false, + "refId": "A" + } + ], + "title": "Ingestion rate", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of HTTP read requests.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "req/s" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 12, + "y": 1 + }, + "id": 130, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_http_requests_total{job=~\"$job\", instance=~\"$instance\", path=~\"/select/.*\"}[$__rate_interval]))", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Read requests", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Total number of available CPUs for all VM components. ", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 18, + "y": 1 + }, + "id": 126, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(vm_available_cpu_cores{job=~\"$job\", instance=~\"$instance\"})", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Available CPU", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of [active time series](https://docs.victoriametrics.com/faq/#what-is-an-active-time-series) with new data points inserted during the last hour. High value may result in ingestion slowdown.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 0, + "y": 4 + }, + "id": 34, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(max_over_time(vm_cache_entries{job=~\"$job\", instance=~\"$instance\", type=\"storage/hour_metric_ids\"}[1h]))", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Active series", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Total amount of used disk space", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 6, + "y": 4 + }, + "id": 35, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": false, + "expr": "sum(vm_data_size_bytes{job=~\"$job_storage\"})", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Disk space usage", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Average disk usage per datapoint.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "decimals": 2, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 12, + "y": 4 + }, + "id": 112, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": true, + "expr": "sum(vm_data_size_bytes{job=~\"$job_storage\"}) / sum(vm_rows{job=~\"$job_storage\", type!~\"indexdb.*\"})", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Bytes per point", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Total size of available memory for all VM components.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 18, + "y": 4 + }, + "id": 128, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(vm_available_memory_bytes{job=~\"$job\", instance=~\"$instance\"})", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Available memory", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "inspect": false, + "minWidth": 50 + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Time" + }, + "properties": [ + { + "id": "custom.hidden", + "value": true + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Value" + }, + "properties": [ + { + "id": "displayName", + "value": "Count" + } + ] + } + ] + }, + "gridPos": { + "h": 5, + "w": 8, + "x": 0, + "y": 7 + }, + "id": 149, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true, + "sortBy": [ + { + "desc": true, + "displayName": "Count" + } + ] + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "sum(vm_app_version{job=~\"$job\", instance=~\"$instance\"}) by(job, short_version)", + "format": "table", + "instant": true, + "range": false, + "refId": "A" + } + ], + "title": "", + "type": "table" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "stepAfter", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 16, + "x": 8, + "y": 7 + }, + "id": 62, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(min_over_time(up{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job)", + "format": "time_series", + "instant": false, + "legendFormat": "{{job}}", + "refId": "A" + } + ], + "title": "Uptime ($job)", + "type": "timeseries" + }, + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 12 + }, + "id": 10, + "panels": [], + "title": "Overview", + "type": "row" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "How many [data samples](https://docs.victoriametrics.com/keyconcepts/#raw-samples) are inserted into cluster per second by protocol before the replication. Check vminsert metrics if there are any issues with ingestion.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineStyle": { + "fill": "solid" + }, + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 13 + }, + "id": 2, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_rows_inserted_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (type) > 0 ", + "interval": "", + "legendFormat": "{{type}}", + "range": true, + "refId": "A" + } + ], + "title": "Datapoints ingestion rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "* `*` - unsupported query path\n* `/write` - insert into VM\n* `/metrics` - query VM system metrics\n* `/query` - query instant values\n* `/query_range` - query over a range of time\n* `/series` - match a certain label set\n* `/label/{}/values` - query a list of label values (variables mostly)", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 13 + }, + "id": 6, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_http_requests_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (path) > 0", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{path}}", + "range": true, + "refId": "A" + } + ], + "title": "Requests rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of [active time series](https://docs.victoriametrics.com/faq/#what-is-an-active-time-series) with new data points inserted during the last hour across all storage nodes. High value may result in ingestion slowdown and high memory usage.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 21 + }, + "id": 12, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(vm_cache_entries{job=~\"$job\", instance=~\"$instance\", type=\"storage/hour_metric_ids\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "Active time series", + "refId": "A" + } + ], + "title": "Active time series ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "The less time it takes is better.\n* `*` - unsupported query path\n* `/write` - insert into VM\n* `/metrics` - query VM system metrics\n* `/query` - query instant values\n* `/query_range` - query over a range of time\n* `/series` - match a certain label set\n* `/label/{}/values` - query a list of label values (variables mostly)", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 21 + }, + "id": 8, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(vm_request_duration_seconds{job=~\"$job\", instance=~\"$instance\", quantile=\"0.99\"}) by (path) > 0", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{path}}", + "range": true, + "refId": "A" + } + ], + "title": "Query duration 0.99 quantile ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "* `*` - unsupported query path\n* `/write` - insert into VM\n* `/metrics` - query VM system metrics\n* `/query` - query instant values\n* `/query_range` - query over a range of time\n* `/series` - match a certain label set\n* `/label/{}/values` - query a list of label values (variables mostly)", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 29 + }, + "id": 52, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_http_request_errors_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job, path) > 0", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{path}} ({{job}})", + "range": true, + "refId": "A" + } + ], + "title": "Requests error rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of logging the messages by their level. Unexpected spike in rate is a good reason to check logs.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "bars", + "fillOpacity": 100, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=203&var-job=${__field.labels.job}&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 29 + }, + "id": 104, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_log_messages_total{job=~\"$job\",instance=~\"$instance\", level!=\"info\"}[$__rate_interval])) by (job, level) > 0", + "format": "time_series", + "hide": false, + "interval": "5m", + "intervalFactor": 1, + "legendFormat": "{{job}} - {{level}}", + "range": true, + "refId": "A" + } + ], + "title": "Logging rate", + "type": "timeseries" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 37 + }, + "id": 46, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Percentage of used RSS memory (resident).\nThe RSS memory shows the amount of memory recently accessed by the application. It includes anonymous memory and data from recently accessed files (aka page cache).\nThe application's performance will significantly degrade when memory usage is close to 100%.\n\nClick on the line and choose Drilldown to show memory usage per instance", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=189&var-job=${__field.labels.job}&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 38 + }, + "id": 66, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n max_over_time(process_resident_memory_bytes{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job\", instance=~\"$instance\"}\n) by(job)", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "RSS memory % usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Share for memory allocated by the process itself. When memory usage reaches 100% it will be likely OOM-killed.\nSafe memory usage % considered to be below 80%\n\nClick on the line and choose Drilldown to show memory usage per instance", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=190&var-job=${__field.labels.job}&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 38 + }, + "id": 138, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n max_over_time(process_resident_memory_anon_bytes{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job\", instance=~\"$instance\"}\n) by(job)", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "RSS anonymous memory % usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=192&var-job=${__field.labels.job}&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 160 + }, + "id": 64, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n rate(process_cpu_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_cpu_cores_available{job=~\"$job\", instance=~\"$instance\"}\n) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "CPU ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of bytes read/write from the storage layer.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "Bps" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/read .*/" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 160 + }, + "id": 122, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(process_io_storage_read_bytes_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "read {{job}}", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(process_io_storage_written_bytes_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "write {{job}}", + "range": true, + "refId": "B" + } + ], + "title": "Disk writes/reads ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the percentage of open file descriptors compared to the limit set in the OS.\nReaching the limit of open files can cause various issues and must be prevented.\n\nSee how to change limits here https://medium.com/@muhammadtriwibowo/set-permanently-ulimit-n-open-files-in-ubuntu-4d61064429a", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 2, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/max.*/" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#C4162A", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 168 + }, + "id": 117, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n max_over_time(process_open_fds{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_max_fds{job=~\"$job\", instance=~\"$instance\"}\n) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{job}}", + "range": true, + "refId": "A" + } + ], + "title": "Open FDs usage % ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of read/write syscalls such as read, pread, write, pwrite.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/read .*/" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 168 + }, + "id": 204, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(process_io_read_syscalls_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "read {{job}}", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(process_io_write_syscalls_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "write {{job}}", + "range": true, + "refId": "B" + } + ], + "title": "Disk write/read calls ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 176 + }, + "id": 68, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(max_over_time(go_goroutines{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Goroutines ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 176 + }, + "id": 119, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(max_over_time(vm_tcplistener_conns{job=~\"$job\", instance=~\"$instance\"}[$__interval])) by(job)", + "interval": "", + "legendFormat": "{{job}}", + "range": true, + "refId": "A" + } + ], + "title": "TCP connections ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 184 + }, + "id": 70, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(max_over_time(process_num_threads{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Threads ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 184 + }, + "id": 120, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_tcplistener_accepts_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job)", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "TCP connections rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the percent of CPU spent on garbage collection.\n\nIf % is high, then CPU usage can be decreased by changing GOGC to higher values. Increasing GOGC value will increase memory usage, and decrease CPU usage.\n\nTry searching for keyword `GOGC` at https://docs.victoriametrics.com/troubleshooting/ ", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 192 + }, + "id": 210, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(\n rate(go_gc_cpu_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]) \n / rate(process_cpu_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n ) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "CPU spent on GC ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the time goroutines have spent in runnable state before actually running. The lower is better.\n\nHigh values or values exceeding the threshold is usually a sign of insufficient CPU resources or CPU throttling. \n\nVerify that service has enough CPU resources. Otherwise, the service could work unreliably with delays in processing.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 0.1 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 192 + }, + "id": 213, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(histogram_quantile(0.99, sum(rate(go_sched_latencies_seconds_bucket{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job, instance, le))) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Go scheduling latency", + "type": "timeseries" + } + ], + "title": "Resource usage ($job)", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 38 + }, + "id": 106, + "panels": [ + { + "description": "", + "fieldConfig": { + "defaults": {}, + "overrides": [] + }, + "gridPos": { + "h": 2, + "w": 24, + "x": 0, + "y": 39 + }, + "id": 211, + "options": { + "code": { + "language": "plaintext", + "showLineNumbers": false, + "showMiniMap": false + }, + "content": "See [Troubleshooting](https://docs.victoriametrics.com/troubleshooting/) docs.", + "mode": "markdown" + }, + "pluginVersion": "11.5.0", + "title": "", + "transparent": true, + "type": "text" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate and total number of new series created over last 24h.\n\nHigh [churn rate](https://docs.victoriametrics.com/faq/#what-is-high-churn-rate) tightly connected with database performance and may result in unexpected OOM's or slow queries. It is recommended to always keep an eye on this metric to avoid unexpected [cardinality](https://docs.victoriametrics.com/keyconcepts/#cardinality) \"explosions\".\n\nThe higher churn rate is, the more resources required to handle it. Consider to keep the churn rate as low as possible.\n\nTo investigate stats about most expensive series use `api/v1/status/tsdb` handler. More details here https://docs.victoriametrics.com/cluster-victoriametrics/#url-format\n\nGood references to read:\n* https://www.robustperception.io/cardinality-is-key\n* https://valyala.medium.com/high-cardinality-tsdb-benchmarks-victoriametrics-vs-timescaledb-vs-influxdb-13e6ee64dd6b", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "new series over 24h" + }, + "properties": [ + { + "id": "custom.axisPlacement", + "value": "right" + }, + { + "id": "custom.axisSoftMin", + "value": 0 + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 104 + }, + "id": 102, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": true, + "expr": "sum(rate(vm_new_timeseries_created_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval]))", + "interval": "", + "legendFormat": "churn rate", + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": true, + "expr": "sum(increase(vm_new_timeseries_created_total{job=~\"$job_storage\", instance=~\"$instance\"}[24h]))", + "hide": false, + "interval": "", + "legendFormat": "new series over 24h", + "refId": "B" + } + ], + "title": "Churn rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "The percentage of [slow inserts](https://docs.victoriametrics.com/faq/#what-is-a-slow-insert) compared to the total ingestion rate. \\n\\nThe lower the better. \\n\\nIn short, slow insert is a cache miss. There are following reasons for slow inserts to go up: \\n* Ingestion of completely new, not seen before time series;\\n* [Re-routing](https://docs.victoriametrics.com/cluster-victoriametrics/#cluster-availability) of series when one or more vmstorage nodes are unavailable;\\n* Not enough memory to maintain big enough caches for the current workload.\\n\\nIf percentage remains high (>10%) during extended periods of time, then it is likely more RAM is needed for optimal handling of the current number of [active time series](https://docs.victoriametrics.com/faq/#what-is-an-active-time-series). \\n\\nSee [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3976#issuecomment-1476883183) for details.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line+area" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "transparent", + "value": null + }, + { + "color": "red", + "value": 0.1 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 104 + }, + "id": 108, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(\n rate(vm_slow_row_inserts_total{job=~\"$job_storage\"}[$__rate_interval]) \n / rate(vm_rows_added_to_storage_total{job=~\"$job_storage\"}[$__rate_interval])\n)", + "interval": "", + "legendFormat": "slow inserts", + "range": true, + "refId": "A" + } + ], + "title": "Slow inserts", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Merge assist happens when vmstorage can't keep up with merging parts. This is usually a sign of overload for vmstorage.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "bars", + "fillOpacity": 100, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 112 + }, + "id": 170, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(increase(vm_assisted_merges_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])) by(type) > 0", + "format": "time_series", + "interval": "5m", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Assisted merges ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows % of slow queries according to `search.logSlowQueryDuration` flag, which is `5s` by default.\n\nThe less value is better.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 112 + }, + "id": 107, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_slow_queries_total{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval]))\n/\nsum(rate(vm_http_requests_total{job=~\"$job_select\", instance=~\"$instance\", path=~\"/select/.*\"}[$__rate_interval]))", + "interval": "", + "legendFormat": "slow queries %", + "range": true, + "refId": "A" + } + ], + "title": "Slow queries % ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the percentage of used cache size from the allowed size by type. \nValues close to 100% show the maximum potential utilization.\nValues close to 0% show that cache is underutilized.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 120 + }, + "id": 144, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n vm_cache_size_bytes{job=~\"$job\", instance=~\"$instance\"} \n /\n vm_cache_size_max_bytes{job=~\"$job\", instance=~\"$instance\"}\n) by(type)", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Cache usage % by type ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows cache miss ratio. Lower is better.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "max": 1, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 120 + }, + "id": 58, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n rate(vm_cache_misses_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n /\n rate(vm_cache_requests_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n) by(type)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Cache miss ratio ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "The value is above 0 when the vmstorage at the given `addr` communicates to the given vminsert node that it cannot accept new data because it is in the read-only mode.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 129 + }, + "id": 142, + "links": [ + { + "targetBlank": true, + "title": "Readonly mode", + "url": "https://docs.victoriametrics.com/cluster-victoriametrics/#readonly-mode" + } + ], + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(vm_rpc_vmstorage_is_read_only{job=~\"$job_insert\", instance=~\"$instance\"}) by(instance, addr) > 0", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{instance}} => {{addr}}", + "range": true, + "refId": "A" + } + ], + "title": "Storage in readonly status for vminsert ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows rate of deduplicated samples during [deduplication](https://docs.victoriametrics.com/#deduplication) or [downsampling](https://docs.victoriametrics.com/#downsampling), according to the configured `-dedup.minScrapeInterval` or `-downsampling.period`.\n\nShould be empty if `-dedup.minScrapeInterval` and `-downsampling.period` isn't set.\n\nDeduplication happens in two places:\n* `type=\"merge\"` - during [background merges](https://docs.victoriametrics.com/#storage) by vmstorages. Deduplication during merges is permanent.\n* `type=\"select\"` - during [read queries](https://docs.victoriametrics.com/keyconcepts/#query-data) by vmselects. Deduplication happens in-flight and has no permanent effect.\n\nThe more samples need to be deduplicated, the higher will be resource usage on vmselects or vmstorages.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 129 + }, + "id": 215, + "links": [], + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_deduplicated_samples_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, type) >0", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Deduplication rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows how many samples were ignored or dropped on insertion due to various reasons:\n* timestamp out of retention period or timestamp in future;\n* invalid metric name;\n* exceeding limit for labels length or number;\n* dropped by [relabeling configuration](https://docs.victoriametrics.com/#relabeling).", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 137 + }, + "id": 135, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(increase(vm_rows_ignored_total{job=~\"$job\", instance=~\"$instance\"}[1h])) by (reason)", + "hide": false, + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(increase(vm_relabel_metrics_dropped_total{job=~\"$job\", instance=~\"$instance\"}[1h]))", + "hide": false, + "instant": false, + "legendFormat": "relabeling", + "range": true, + "refId": "B" + } + ], + "title": "Samples dropped for last 1h ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of partial results served by `vmselects`.\n\n`vmselect` continues serving queries if at least a single vmstorage nodes is available. It marks responses as `partial` for queries served from the remaining healthy vmstorage nodes.\n\nIf you prefer consistency over availability then run vmselect nodes with `-search.denyPartialResponse` cmd-line flag. In this case vmselect returns an error if at least a single vmstorage node is unavailable.\n\nSee more at [cluster availability](https://docs.victoriametrics.com/cluster-victoriametrics/#cluster-availability).", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "stepAfter", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 137 + }, + "id": 217, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(increase(vm_partial_results_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) > 0", + "format": "time_series", + "instant": false, + "legendFormat": "partial results", + "refId": "A" + } + ], + "title": "Partial query results ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "inspect": false + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Value" + }, + "properties": [ + { + "id": "custom.hidden", + "value": true + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Time" + }, + "properties": [ + { + "id": "custom.hidden", + "value": true + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 145 + }, + "id": 183, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true, + "sortBy": [ + { + "desc": true, + "displayName": "job" + } + ] + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "sum(flag{is_set=\"true\", job=~\"$job\", instance=~\"$instance\"}) by(job, instance, name, value)", + "format": "table", + "instant": true, + "legendFormat": "__auto", + "range": false, + "refId": "A" + } + ], + "title": "Non-default flags", + "transformations": [ + { + "id": "groupBy", + "options": { + "fields": { + "instance": { + "aggregations": ["uniqueValues"], + "operation": "aggregate" + }, + "job": { + "aggregations": [], + "operation": "groupby" + }, + "name": { + "aggregations": [], + "operation": "groupby" + }, + "value": { + "aggregations": [], + "operation": "groupby" + } + } + } + } + ], + "type": "table" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of restarts per job. The chart can be useful to identify periodic process restarts and correlate them with potential issues or anomalies. Normally, processes shouldn't restart unless restart was inited by user. The reason of restarts should be figured out by checking the logs of each specific service. ", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "stepAfter", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 145 + }, + "id": 214, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(changes(vm_app_start_timestamp{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "format": "time_series", + "instant": false, + "legendFormat": "{{job}}", + "refId": "A" + } + ], + "title": "Restarts ($job)", + "type": "timeseries" + } + ], + "title": "Troubleshooting", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 39 + }, + "id": 48, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows number of pushed and sent rows. \n* `Pushed rows` - rows added to internal inserter buffers before send\n* `Sent rows` - successfully transmitted rows to storage nodes\n\nPlease note, it could be that `Sent > Pushed` because of the replication factor.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 1331 + }, + "id": 76, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_rpc_rows_pushed_total{job=~\"$job\",instance=~\"$instance\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "Pushed", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_rpc_rows_sent_total{job=~\"$job\",instance=~\"$instance\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "Sent", + "range": true, + "refId": "E" + } + ], + "title": "Rows ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Rate of RPC errors by type:\n* `Connection` - the number of connection errors to vmstorage node\n* `Dial` - the number of dial errors to vmstorage node.\n* `Handshake` - the number of handshake errors to vmstorage node\n* `Rerouted` - errors appeared during rerouting of rows from un-healthy storage node to a healthy one.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 1331 + }, + "id": 86, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_rpc_connection_errors_total{job=~\"$job\",instance=~\"$instance\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "Connection", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(rate(vm_rpc_dial_errors_total{job=~\"$job\",instance=~\"$instance\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "Dial", + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(rate(vm_rpc_handshake_errors_total{job=~\"$job\",instance=~\"$instance\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "Handshake", + "refId": "E" + } + ], + "title": "RPC errors ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "The number of rows rerouted to the vmstorage node from other nodes when they were unhealthy.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "rps" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1340 + }, + "id": 80, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_rpc_rows_rerouted_to_here_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(addr) > 0", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Rows ($instance) rerouted to ", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "The number of rows rerouted from the vmstorage node to healthy nodes when the given node was unhealthy.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "rps" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1340 + }, + "id": 78, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": true, + "expr": "sum(rate(vm_rpc_rows_rerouted_from_here_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(addr) > 0", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{addr}}", + "refId": "A" + } + ], + "title": "Rows ($instance) rerouted from", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "The number of rows or bytes that vminsert internal buffer contains at the moment.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "bytes" + }, + "properties": [ + { + "id": "unit", + "value": "bytes" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1348 + }, + "id": 82, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(vm_rpc_buf_pending_bytes{job=~\"$job\", instance=~\"$instance\"})", + "legendFormat": "bytes", + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(vm_rpc_rows_pending{job=~\"$job\", instance=~\"$instance\"})", + "legendFormat": "rows", + "refId": "B" + } + ], + "title": "Pending", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Network usage by internal VictoriaMetrics RPC protocol", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bps" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1348 + }, + "id": 74, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcpdialer_written_bytes_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) * 8", + "legendFormat": "network usage", + "range": true, + "refId": "A" + } + ], + "title": "RPC network usage ($instance)", + "type": "timeseries" + } + ], + "title": "Interconnection ($job)", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 40 + }, + "id": 24, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of datapoints ingested into storage nodes per second. This metric doesn't show all stored datapoints since some of them may be dropped because of wrong timestamps or decode errors.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 41 + }, + "id": 100, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_vminsert_metrics_read_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])) ", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "ingestion rate", + "range": true, + "refId": "A" + } + ], + "title": "Ingestion rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the approx time needed to reach 100% of disk capacity for at least one vmstorage node based on the following params:\n* free disk space (after -storage.minFreeDiskSpaceBytes);\n* row ingestion rate;\n* compression.\n\nNote: this panel doesn't account for deduplication process.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=196&${__url_time_range}&${__all_variables}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 41 + }, + "id": 113, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "min"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "min((vm_free_disk_space_bytes{job=~\"$job_storage\", instance=~\"$instance\"} - vm_free_disk_space_limit_bytes{job=~\"$job_storage\", instance=~\"$instance\"}) \n/ \nignoring(path) (\n rate(vm_rows_added_to_storage_total{job=~\"$job_storage\", instance=~\"$instance\"}[1d])\n * scalar(\n sum(vm_data_size_bytes{job=~\"$job_storage\", instance=~\"$instance\", type!~\"indexdb.*\"})\n / \n sum(vm_rows{job=~\"$job_storage\", instance=~\"$instance\", type!~\"indexdb.*\"})\n )\n))", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "min ETA", + "range": true, + "refId": "A" + } + ], + "title": "Storage full ETA ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=192&var-job=$job_storage&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 49 + }, + "id": 151, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "max(\n rate(process_cpu_seconds_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_cpu_cores_available{job=~\"$job_storage\", instance=~\"$instance\"}\n)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "max", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "min(\n rate(process_cpu_seconds_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_cpu_cores_available{job=~\"$job_storage\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "min", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "quantile(0.5,\n rate(process_cpu_seconds_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_cpu_cores_available{job=~\"$job_storage\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "median", + "range": true, + "refId": "C" + } + ], + "title": "CPU usage % ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Percentage of used memory (resident).\nThe application's performance will significantly degrade when memory usage is close to 100%.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=189&var-job=$job_storage&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 49 + }, + "id": 167, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n max_over_time(process_resident_memory_bytes{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job_storage\", instance=~\"$instance\"}\n)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "max", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "min(\n max_over_time(process_resident_memory_bytes{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job_storage\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "min", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "quantile(0.5,\n max_over_time(process_resident_memory_bytes{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job_storage\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "median", + "range": true, + "refId": "C" + } + ], + "title": "Memory usage % ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows how many ongoing insertions (not API /write calls) on disk are taking place, where:\n* `max` - equal to number of CPUs;\n* `current` - current number of goroutines busy with inserting rows into underlying storage.\n\nEvery successful API /write call results into flush on disk. The `max` is an internal limit and can't be changed. It is always equal to the number of CPUs. \n\nWhen `current` hits `max` constantly, it means storage is overloaded and requires more CPU (see CPU usage) or disks with more IOPS (see disk writes and reads panels in Resource Usage row).", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "max" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#C4162A", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 56 + }, + "id": 212, + "links": [ + { + "targetBlank": true, + "title": "Related discussion", + "url": "https://github.com/VictoriaMetrics/VictoriaMetrics/issues/632" + } + ], + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n max_over_time(vm_concurrent_insert_current{job=~\"$job_storage\", \n instance=~\"$instance\"}[$__rate_interval])\n)", + "interval": "", + "legendFormat": "current", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "min(vm_concurrent_insert_capacity{job=~\"$job_storage\", instance=~\"$instance\"})", + "hide": false, + "interval": "", + "legendFormat": "max", + "range": true, + "refId": "B" + } + ], + "title": "Concurrent flushes on disk ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of concurrently executed read requests, where:\n* `max` - equal to `-search.maxConcurrentRequest` cmd-line flag;\n* `current` - current number of concurrent select requests executed by storage.\n\nEach concurrent select request requires RAM, CPU and disk IO resources during processing. Too many concurrent requests could result into OOM exceptions.\n\nWhen `current` hits `max` constantly, it means storage is overloaded and requires more CPU (see CPU usage) or disks with more IOPS (see disk writes and reads panels in Resource Usage row).\n\nSee more at https://docs.victoriametrics.com/cluster-victoriametrics/#resource-usage-limits", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "max" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#C4162A", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 56 + }, + "id": 133, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n max_over_time(vm_vmselect_concurrent_requests_current{job=~\"$job_storage\", \n instance=~\"$instance\"}[$__rate_interval])\n)", + "interval": "", + "legendFormat": "current", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "min(vm_vmselect_concurrent_requests_capacity{job=~\"$job_storage\", instance=~\"$instance\"})", + "hide": false, + "interval": "", + "legendFormat": "max", + "range": true, + "refId": "B" + } + ], + "title": "Concurrent selects ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "The max number of on-going merges across storage nodes.\n The drastic change in number of merges could be a sign of on-going deduplication/downsampling activity.\n It is expected to have high numbers for `storage/small` metric.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 64 + }, + "id": 54, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(max_over_time(vm_active_merges{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])) by(type)", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Active merges ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "The number of rows merged per second by storage nodes. Merge speed depends on available CPU and disk IO bandwidth.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 64 + }, + "id": 55, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_rows_merged_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])) by(type)", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Merge speed", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the percentage of used disk space. It is recommended to have at least 20% of free disk space for the best performance.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=200&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 72 + }, + "id": 20, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(\n sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance) /\n (\n sum(vm_free_disk_space_bytes{job=~\"$job\", instance=~\"$instance\"}-vm_free_disk_space_limit_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance) +\n sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance)\n ) \n)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "max", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "min(\n sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance) /\n (\n sum(vm_free_disk_space_bytes{job=~\"$job\", instance=~\"$instance\"}-vm_free_disk_space_limit_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance) +\n sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance)\n ) \n)", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "min", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "quantile(0.5,\n sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance) /\n (\n sum(vm_free_disk_space_bytes{job=~\"$job\", instance=~\"$instance\"}-vm_free_disk_space_limit_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance) +\n sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance)\n ) \n)", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "median", + "range": true, + "refId": "C" + } + ], + "title": "Disk space usage % ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "The max number of data parts of LSM tree across all storage nodes in the last [partition](https://docs.victoriametrics.com/#storage).\nIncrease in number of parts (the hard limit is 512) is an evidence of slow merge performance - check the resource utilization.\n* `indexdb` - inverted index\n* `storage/small` - recently added parts of data ingested into storage (hot data)\n* `storage/big` - small parts gradually merged into bigger parts (cold data)", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 72 + }, + "id": 22, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(vm_parts{job=~\"$job_storage\", instance=~\"$instance\", type=~\"indexdb.*\"}) by(type)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(vm_last_partition_parts{job=~\"$job_storage\", instance=~\"$instance\"}) by(type)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "B" + } + ], + "title": "LSM parts max by type ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the percentage of used disk space by type: datapoints or indexdb. Normally, indexdb takes much less space comparing to datapoints. But with high [churn rate](https://docs.victoriametrics.com/faq/#what-is-high-churn-rate) the size of the indexdb could grow significantly.\n\nThe sum of the % can be > 100% since panel shows max % per-job and per-instance. It means different instance can have different ratio between datapoints and indexdb size.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=201&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 80 + }, + "id": 202, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(\n sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\", type=~\"indexdb.*\"}) by(job, instance)\n / \n sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance)\n)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "indexdb", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(\n sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\", type!~\"indexdb.*\"}) by(job, instance)\n / \n sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance)\n)", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "datapoints", + "range": true, + "refId": "B" + } + ], + "title": "Disk space usage % by type ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "How many datapoints are in RAM queue waiting to be written into storage. The number of pending data points should be in the range from 0 to `3*`, since VictoriaMetrics pushes pending data to persistent storage every two seconds. The index datapoints value in general is much lower.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "pending index entries" + }, + "properties": [ + { + "id": "unit", + "value": "none" + }, + { + "id": "decimals", + "value": 3 + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 80 + }, + "id": 14, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(max_over_time(vm_pending_rows{job=~\"$job_storage\", instance=~\"$instance\", type=\"storage\"}[$__rate_interval]))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "pending datapoints", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(max_over_time(vm_pending_rows{job=~\"$job_storage\", instance=~\"$instance\", type=\"indexdb\"}[$__rate_interval]))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "pending index entries", + "range": true, + "refId": "B" + } + ], + "title": "Pending datapoints ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows network usage by vmstorage services.\n* Writes show traffic sent to clients.\n* Reads show traffic received from clients.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bps" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/read.*/" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 88 + }, + "id": 206, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcplistener_read_bytes_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])) by(name) * 8 > 0", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "read from {{name}}", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcplistener_written_bytes_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])) by(name) * 8 > 0", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "write to {{name}}", + "range": true, + "refId": "B" + } + ], + "title": "Network usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Value > 0 means vmstorage is in readonly mode.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 88 + }, + "id": 141, + "links": [ + { + "targetBlank": true, + "title": "Readonly mode", + "url": "https://docs.victoriametrics.com/cluster-victoriametrics/#readonly-mode" + } + ], + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "vm_storage_is_read_only{job=~\"$job_storage\", instance=~\"$instance\"} > 0", + "interval": "", + "legendFormat": "{{ instance }}", + "range": true, + "refId": "A" + } + ], + "title": "Readonly mode", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the max number of existing [snapshots](https://docs.victoriametrics.com/#how-to-work-with-snapshots) across vmstorages.\n\nRecently made snapshots do not occupy disk space. But with time, as snapshots become older, they start to occupy more and more disk space.\n\nIt is recommended deleting old snapshots when they are no longer needed to free up disk space.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 96 + }, + "id": 216, + "links": [], + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "11.5.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(vm_snapshots{job=~\"$job_storage\", instance=~\"$instance\"})", + "interval": "", + "legendFormat": "snapshots", + "range": true, + "refId": "A" + } + ], + "title": "Number of snapshots", + "type": "timeseries" + } + ], + "title": "vmstorage ($instance)", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 41 + }, + "id": 42, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Request rate accepted by vmselect nodes", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1344 + }, + "id": 92, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_http_requests_total{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (path) > 0", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Requests rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the max number of concurrent selects across instances.\n* `max` limit can be configured via `search.maxConcurrentRequests` flag\n* `current` shows the current number of goroutines busy with processing requests\n\nWhen `current` hits `max` constantly, it means one or more vmselect nodes are overloaded with number of requests. If you observe that CPU for vmselects is saturated, consider adding more vmselect replicas or increase CPU resources. If CPU panel shows a plenty of free resources - try increasing `search.maxConcurrentRequests`.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "max" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#C4162A", + "mode": "fixed" + } + }, + { + "id": "custom.fillOpacity", + "value": 0 + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1344 + }, + "id": 95, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(max_over_time(vm_concurrent_select_current{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) ", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "current", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "min(vm_concurrent_select_capacity{job=~\"$job_select\", instance=~\"$instance\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "max", + "refId": "B" + } + ], + "title": "Concurrent selects ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=192&var-job=$job_select&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1352 + }, + "id": 163, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n rate(process_cpu_seconds_total{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_cpu_cores_available{job=~\"$job_select\", instance=~\"$instance\"}\n)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "max", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "min(\n rate(process_cpu_seconds_total{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_cpu_cores_available{job=~\"$job_select\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "min", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "quantile(0.5,\n rate(process_cpu_seconds_total{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_cpu_cores_available{job=~\"$job_select\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "median", + "range": true, + "refId": "C" + } + ], + "title": "CPU usage % ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=189&var-job=$job_select&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1352 + }, + "id": 165, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n max_over_time(process_resident_memory_bytes{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job_select\", instance=~\"$instance\"}\n)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "max", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "min(\n max_over_time(process_resident_memory_bytes{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job_select\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "min", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "quantile(0.5,\n max_over_time(process_resident_memory_bytes{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job_select\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "median", + "range": true, + "refId": "C" + } + ], + "title": "Memory usage % ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "99th percentile of the number of time series read per query.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 2, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1360 + }, + "id": 178, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(histogram_quantile(0.99, sum(rate(vm_series_read_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "series", + "range": true, + "refId": "A" + } + ], + "title": "Series read per query ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "99th percentile of number of [data samples](https://docs.victoriametrics.com/keyconcepts/#raw-samples) read per queried time series.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 2, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1360 + }, + "id": 180, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(histogram_quantile(0.99, sum(rate(vm_rows_read_per_series_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "datapoints", + "range": true, + "refId": "A" + } + ], + "title": "Datapoints read per series ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "99th percentile of number of [data samples](https://docs.victoriametrics.com/keyconcepts/#raw-samples) read per query.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 2, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1368 + }, + "id": 179, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(histogram_quantile(0.99, sum(rate(vm_rows_read_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "datapoints", + "range": true, + "refId": "A" + } + ], + "title": "Datapoints read per query ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "99th percentile of number of [data samples](https://docs.victoriametrics.com/keyconcepts/#raw-samples) scanner per query.\n\nThis number can exceed number of DatapointsReadPerQuery if `step` query arg passed to [/api/v1/query_range](https://prometheus.io/docs/prometheus/latest/querying/api/#range-queries) is smaller than the lookbehind window set in square brackets of [rollup function](https://docs.victoriametrics.com/metricsql/#rollup-functions). For example, if `increase(some_metric[1h])` is executed with the `step=5m`, then the same [data samples](https://docs.victoriametrics.com/keyconcepts/#raw-samples) on a hour time range are scanned `1h/5m=12` times. See [this article](https://valyala.medium.com/how-to-optimize-promql-and-metricsql-queries-85a1b75bf986) for details.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 2, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1368 + }, + "id": 181, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(histogram_quantile(0.99, sum(rate(vm_rows_scanned_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "datapoints", + "range": true, + "refId": "A" + } + ], + "title": "Datapoints scanned per query ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows network usage between vmselects and clients, such as vmalert, Grafana, vmui, etc.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bps" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/read.*/" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1376 + }, + "id": 93, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcplistener_read_bytes_total{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) * 8 > 0", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "read from client", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcplistener_written_bytes_total{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) * 8 > 0", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "write to client", + "range": true, + "refId": "B" + } + ], + "title": "Network usage: clients ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows network usage between vmselects and vmstorages.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bps" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/read.*/" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1376 + }, + "id": 207, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcpdialer_read_bytes_total{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) * 8 > 0", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "read from vmstorage", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcpdialer_written_bytes_total{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) * 8 > 0", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "write to vmstorage", + "range": true, + "refId": "B" + } + ], + "title": "Network usage: vmstorage ($instance)", + "type": "timeseries" + } + ], + "title": "vmselect ($instance)", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 42 + }, + "id": 40, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "* `*` - unsupported query path\n* `/write` - insert into VM\n* `/metrics` - query VM system metrics\n* `/query` - query instant values\n* `/query_range` - query over a range of time\n* `/series` - match a certain label set\n* `/label/{}/values` - query a list of label values (variables mostly)", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1345 + }, + "id": 97, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_http_requests_total{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval])) by (path) > 0", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Requests rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the max number of ongoing insertions.\n* `max` - equal to number of CPU * 2 by default. May be configured with `maxConcurrentInserts` flag;\n* `current` - current number of goroutines busy with processing requests.\n\n`-maxConcurrentInserts` limits the number of insert requests which may be actively processed at any given point in time. All the other insert requests are queued for up to `-insert.maxQueueDuration` in the hope they will get a chance to be processed. This queue is used mostly for absorbing spikes for incoming insert request rate.\n\nWhen `current` hits `max` constantly, it means vminsert node is overloaded and requires more CPU or higher limits.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "max" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#C4162A", + "mode": "fixed" + } + }, + { + "id": "custom.fillOpacity", + "value": 0 + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1345 + }, + "id": 99, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(max_over_time(vm_concurrent_insert_current{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval]))", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "current", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": true, + "expr": "min(vm_concurrent_insert_capacity{job=~\"$job_insert\", instance=~\"$instance\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "max", + "refId": "B" + } + ], + "title": "Concurrent inserts ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=192&var-job=$job_insert&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1353 + }, + "id": 185, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n rate(process_cpu_seconds_total{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_cpu_cores_available{job=~\"$job_insert\", instance=~\"$instance\"}\n)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "max", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "min(\n rate(process_cpu_seconds_total{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_cpu_cores_available{job=~\"$job_insert\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "min", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "quantile(0.5,\n rate(process_cpu_seconds_total{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_cpu_cores_available{job=~\"$job_insert\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "median", + "range": true, + "refId": "C" + } + ], + "title": "CPU usage % ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/oS7Bi_0Wz?viewPanel=189&var-job=$job_insert&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1353 + }, + "id": 187, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n max_over_time(process_resident_memory_bytes{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job_insert\", instance=~\"$instance\"}\n)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "max", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "min(\n max_over_time(process_resident_memory_bytes{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job_insert\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "min", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "quantile(0.5,\n max_over_time(process_resident_memory_bytes{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job_insert\", instance=~\"$instance\"}\n)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "median", + "range": true, + "refId": "C" + } + ], + "title": "Memory usage % ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the saturation level of connection between vminsert and vmstorage components. \n\nIf the threshold of 0.9sec is reached, then the connection is saturated by more than 90% and vminsert won't be able to keep up. This usually means that either vminsert or vmstorage nodes are struggling with the load. Verify CPU/mem saturation of both components and network saturation between them.\nIf vminsert resources are saturated - consider adding more resources or scale vminserts horizontally.\n\nIf vminsert resources and network are fine, check vmstorage metrics for anomalies.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line+area" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "transparent" + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1361 + }, + "id": 139, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(rate(vm_rpc_send_duration_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(addr)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Storage connection saturation ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows when vmstorage node is unreachable for vminsert.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1361 + }, + "id": 114, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "vm_rpc_vmstorage_is_reachable{job=~\"$job\", instance=~\"$instance\"} != 1", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{instance}} => {{addr}}", + "range": true, + "refId": "A" + } + ], + "title": "Storage reachability ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows network usage between vminserts and clients, such as vmagent, Prometheus, or any other client pushing metrics to vminsert.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bps" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/read.*/" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1369 + }, + "id": 208, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcplistener_read_bytes_total{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval])) * 8 > 0", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "read from client", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcplistener_written_bytes_total{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval])) * 8 > 0", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "write to client", + "range": true, + "refId": "B" + } + ], + "title": "Network usage: clients ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows network usage between vminserts and vmstorages.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bps" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/read.*/" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1369 + }, + "id": 209, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcpdialer_read_bytes_total{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval])) * 8 > 0", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "read from vmstorage", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcpdialer_written_bytes_total{job=~\"$job_insert\", instance=~\"$instance\"}[$__rate_interval])) * 8 > 0", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "write to vmstorage", + "range": true, + "refId": "B" + } + ], + "title": "Network usage: vmstorage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 2, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1377 + }, + "id": 88, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(histogram_quantile(0.99, sum(increase(vm_rows_per_insert_bucket{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "max", + "range": true, + "refId": "A" + } + ], + "title": "Rows per insert ($instance)", + "type": "timeseries" + } + ], + "title": "vminsert ($instance)", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 43 + }, + "id": 194, + "panels": [ + { + "fieldConfig": { + "defaults": {}, + "overrides": [] + }, + "gridPos": { + "h": 2, + "w": 24, + "x": 0, + "y": 1394 + }, + "id": 198, + "options": { + "code": { + "language": "plaintext", + "showLineNumbers": false, + "showMiniMap": false + }, + "content": "Drilldown row is used by other panels on the dashboard to show more detailed metrics per-instance.", + "mode": "markdown" + }, + "pluginVersion": "9.2.7", + "title": "", + "transparent": true, + "type": "text" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the used memory (resident).\nThe application's performance will significantly degrade when memory usage is close to 100%.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + } + ] + }, + "unit": "decbytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1396 + }, + "id": 189, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max_over_time(process_resident_memory_bytes{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])", + "interval": "", + "legendFormat": "{{instance}} ({{job}})", + "range": true, + "refId": "A" + } + ], + "title": "RSS memory usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "RSS share for memory allocated by the process itself. This share cannot be freed by the OS, so it must be taken into account by OOM killer.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + } + ] + }, + "unit": "decbytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1396 + }, + "id": 190, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max_over_time(process_resident_memory_anon_bytes{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])", + "interval": "", + "legendFormat": "{{instance}} ({{job}})", + "range": true, + "refId": "A" + } + ], + "title": "RSS anonymous memory usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 1404 + }, + "id": 192, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "sum(rate(process_cpu_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{instance}} ({{job}})", + "range": true, + "refId": "A" + } + ], + "title": "CPU usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the approx time needed to reach 100% of disk capacity based on the following params:\n* free disk space (after -storage.minFreeDiskSpaceBytes);\n* row ingestion rate;\n* compression.\n\nNote: this panel doesn't account for deduplication process.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 1404 + }, + "id": 196, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "min"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "(vm_free_disk_space_bytes{job=~\"$job_storage\", instance=~\"$instance\"}-vm_free_disk_space_limit_bytes{job=~\"$job_storage\", instance=~\"$instance\"}) \n/ \nignoring(path) (\n rate(vm_rows_added_to_storage_total{job=~\"$job_storage\", instance=~\"$instance\"}[1d])\n * scalar(\n sum(vm_data_size_bytes{job=~\"$job_storage\", instance=~\"$instance\", type!~\"indexdb.*\"})\n / \n sum(vm_rows{job=~\"$job_storage\", instance=~\"$instance\", type!~\"indexdb.*\"})\n )\n)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{instance}}", + "range": true, + "refId": "A" + } + ], + "title": "Storage full ETA ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the percentage of used disk space. It is recommended to have at least 20% of free disk space for the best performance.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + } + ] + }, + "unit": "decbytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1411 + }, + "id": 200, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "", + "range": true, + "refId": "A" + } + ], + "title": "Disk space usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + } + ] + }, + "unit": "decbytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 1411 + }, + "id": 201, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\", type=~\"indexdb.*\"}) by(job, instance)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{job}}:{{instance}} (indexdb)", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(vm_data_size_bytes{job=~\"$job\", instance=~\"$instance\", type!~\"indexdb.*\"}) by(job, instance)", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{job}}:{{instance}} (datapoints)", + "range": true, + "refId": "B" + } + ], + "title": "Disk space usage by type ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of logging the messages by their level. Unexpected spike in rate is a good reason to check logs.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "bars", + "fillOpacity": 100, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 1419 + }, + "id": 203, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_log_messages_total{job=~\"$job\",instance=~\"$instance\", level!=\"info\"}[$__rate_interval])) by (job, instance, level, location) > 0", + "format": "time_series", + "hide": false, + "interval": "5m", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Logging rate", + "type": "timeseries" + } + ], + "title": "Drilldown", + "type": "row" + } + ], + "preload": false, + "refresh": "", + "schemaVersion": 40, + "tags": [], + "templating": { + "list": [ + { + "current": { + "text": "VictoriaMetrics - cluster", + "value": "PAF93674D0B4E9963" + }, + "includeAll": false, + "name": "ds", + "options": [], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "current": {}, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "definition": "label_values(vm_app_version{version=~\"^vm(insert|select|storage).*\"}, job)", + "includeAll": true, + "multi": true, + "name": "job", + "options": [], + "query": { + "query": "label_values(vm_app_version{version=~\"^vm(insert|select|storage).*\"}, job)", + "refId": "VictoriaMetrics-job-Variable-Query" + }, + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "current": {}, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "definition": "label_values(vm_app_version{job=~\"$job\", version=~\"^vminsert.*\"}, job)", + "hide": 2, + "includeAll": true, + "multi": true, + "name": "job_insert", + "options": [], + "query": { + "query": "label_values(vm_app_version{job=~\"$job\", version=~\"^vminsert.*\"}, job)", + "refId": "StandardVariableQuery" + }, + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "current": {}, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "definition": "label_values(vm_app_version{job=~\"$job\", version=~\"^vmselect.*\"}, job)", + "hide": 2, + "includeAll": true, + "multi": true, + "name": "job_select", + "options": [], + "query": { + "query": "label_values(vm_app_version{job=~\"$job\", version=~\"^vmselect.*\"}, job)", + "refId": "StandardVariableQuery" + }, + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "current": {}, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "definition": "label_values(vm_app_version{job=~\"$job\", version=~\"^vmstorage.*\"}, job)", + "hide": 2, + "includeAll": true, + "multi": true, + "name": "job_storage", + "options": [], + "query": { + "query": "label_values(vm_app_version{job=~\"$job\", version=~\"^vmstorage.*\"}, job)", + "refId": "StandardVariableQuery" + }, + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "allValue": ".*", + "current": {}, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "definition": "label_values(vm_app_version{job=~\"$job\"}, instance)", + "includeAll": true, + "multi": true, + "name": "instance", + "options": [], + "query": { + "query": "label_values(vm_app_version{job=~\"$job\"}, instance)", + "refId": "VictoriaMetrics-instance-Variable-Query" + }, + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "baseFilters": [], + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "filters": [], + "name": "adhoc", + "type": "adhoc" + } + ] + }, + "time": { + "from": "now-3h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ] + }, + "timezone": "", + "title": "VictoriaMetrics - cluster", + "uid": "oS7Bi_0Wz", + "version": 1, + "weekStart": "", + "gnetId": 11176 +} diff --git a/rootfs/usr/share/grafana/dashboards/drycc_victoriametrics_vmagent.json b/rootfs/usr/share/grafana/dashboards/drycc_victoriametrics_vmagent.json new file mode 100644 index 0000000..46a825b --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/drycc_victoriametrics_vmagent.json @@ -0,0 +1,7249 @@ +{ + "__inputs": [], + "__elements": {}, + "__requires": [ + { + "type": "grafana", + "id": "grafana", + "name": "Grafana", + "version": "10.4.2" + }, + { + "type": "datasource", + "id": "prometheus", + "name": "Prometheus", + "version": "1.0.0" + }, + { + "type": "panel", + "id": "stat", + "name": "Stat", + "version": "" + }, + { + "type": "panel", + "id": "table", + "name": "Table", + "version": "" + }, + { + "type": "panel", + "id": "text", + "name": "Text", + "version": "" + }, + { + "type": "panel", + "id": "timeseries", + "name": "Time series", + "version": "" + } + ], + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "target": { + "limit": 100, + "matchAny": false, + "tags": [], + "type": "dashboard" + }, + "type": "dashboard" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "enable": true, + "expr": "sum(vm_app_version{job=~\"$job\", instance=~\"$instance\"}) by(short_version) unless (sum(vm_app_version{job=~\"$job\", instance=~\"$instance\"} offset $__interval) by(short_version))", + "hide": true, + "iconColor": "dark-blue", + "name": "version", + "textFormat": "{{short_version}}", + "titleFormat": "Version change" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "enable": true, + "expr": "sum(changes(vm_app_start_timestamp{job=~\"$job\", instance=~\"$instance\"}[$__interval])) by(job, instance)", + "iconColor": "dark-yellow", + "name": "restarts", + "textFormat": "{{job}}:{{instance}} restarted" + } + ] + }, + "description": "Overview for VictoriaMetrics vmagent v1.102.0 or higher", + "editable": false, + "fiscalYearStartMonth": 0, + "graphTooltip": 1, + "id": null, + "links": [ + { + "icon": "doc", + "tags": [], + "targetBlank": true, + "title": "vmagent wiki", + "tooltip": "", + "type": "link", + "url": "https://docs.victoriametrics.com/vmagent/" + }, + { + "icon": "external link", + "tags": [], + "targetBlank": true, + "title": "Found a bug?", + "type": "link", + "url": "https://github.com/VictoriaMetrics/VictoriaMetrics/issues" + }, + { + "icon": "external link", + "tags": [], + "targetBlank": true, + "title": "New releases", + "type": "link", + "url": "https://github.com/VictoriaMetrics/VictoriaMetrics/releases" + } + ], + "liveNow": false, + "panels": [ + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 105, + "panels": [], + "title": "Stats", + "type": "row" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of [samples](https://docs.victoriametrics.com/keyconcepts/#raw-samples) scraped from configured targets.", + "fieldConfig": { + "defaults": { + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 0, + "y": 1 + }, + "id": 103, + "links": [ + { + "title": "Troubleshooting", + "url": "https://docs.victoriametrics.com/vmagent/#troubleshooting" + } + ], + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_promscrape_scraped_samples_sum{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]))", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Samples scraped/s", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of ingested [samples](https://docs.victoriametrics.com/keyconcepts/#raw-samples)", + "fieldConfig": { + "defaults": { + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 4, + "y": 1 + }, + "id": 102, + "links": [ + { + "title": "Troubleshooting", + "url": "https://docs.victoriametrics.com/vmagent/#troubleshooting" + } + ], + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmagent_rows_inserted_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]))", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Samples ingested/s", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of targets scraped per second.", + "fieldConfig": { + "defaults": { + "decimals": 1, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 8, + "y": 1 + }, + "id": 134, + "options": { + "colorMode": "value", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_promscrape_scrapes_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) ", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Targets scraped/s", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows total number of all configured scrape targets in state `up` or `down`.\n\nSee `http://vmagent-host:8429/targets` to get list of all targets. \n", + "fieldConfig": { + "defaults": { + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "down" + }, + "properties": [ + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 1 + } + ] + } + } + ] + } + ] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 12, + "y": 1 + }, + "id": 72, + "links": [ + { + "title": "Troubleshooting", + "url": "https://docs.victoriametrics.com/vmagent/#troubleshooting" + } + ], + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(vm_promscrape_targets{job=~\"$job\", instance=~\"$instance\", status=\"up\"})", + "interval": "", + "legendFormat": "up", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(vm_promscrape_targets{job=~\"$job\", instance=~\"$instance\", status=\"down\"})", + "hide": false, + "interval": "", + "legendFormat": "down", + "range": true, + "refId": "B" + } + ], + "title": "Scrape targets", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows number of generated error messages in logs over last 30m. Non-zero value may be a sign of connectivity or missconfiguration errors.", + "fieldConfig": { + "defaults": { + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 1 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 16, + "y": 1 + }, + "id": 16, + "links": [ + { + "targetBlank": true, + "title": "Troubleshooting", + "url": "https://docs.victoriametrics.com/vmagent/#troubleshooting" + } + ], + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(increase(vm_log_messages_total{job=~\"$job\", instance=~\"$instance\", level!=\"info\"}[30m]))", + "interval": "", + "legendFormat": "", + "refId": "A" + } + ], + "title": "Log errors (30m)", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Persistent queue size shows size of pending samples in bytes which hasn't been flushed to remote storage yet. \nIncreasing of value might be a sign of connectivity issues. In such cases, vmagent starts to flush pending data on disk with attempt to send it later once connection is restored.", + "fieldConfig": { + "defaults": { + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 10485760 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 20, + "y": 1 + }, + "id": 56, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(vm_persistentqueue_bytes_pending{job=~\"$job\", instance=~\"$instance\"})", + "interval": "", + "legendFormat": "", + "refId": "A" + } + ], + "title": "Persistent queue size", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "inspect": false, + "minWidth": 50 + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Time" + }, + "properties": [ + { + "id": "custom.hidden", + "value": true + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Value" + }, + "properties": [ + { + "id": "displayName", + "value": "Count" + } + ] + } + ] + }, + "gridPos": { + "h": 5, + "w": 8, + "x": 0, + "y": 4 + }, + "id": 101, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "sum(vm_app_version{job=~\"$job\", instance=~\"$instance\"}) by(job, short_version)", + "format": "table", + "instant": true, + "range": false, + "refId": "A" + } + ], + "type": "table" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "stepAfter", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 16, + "x": 8, + "y": 4 + }, + "id": 13, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(up{job=~\"$job\", instance=~\"$instance\"}) by (job)", + "format": "time_series", + "instant": false, + "interval": "", + "legendFormat": "__auto", + "refId": "A" + } + ], + "title": "Uptime", + "type": "timeseries" + }, + { + "collapsed": false, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 9 + }, + "id": 24, + "panels": [], + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "refId": "A" + } + ], + "title": "Overview", + "type": "row" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows in/out [samples](https://docs.victoriametrics.com/keyconcepts/#raw-samples) rate including push and pull models. \n\nThe out-rate could be different to in-rate because of replication or additional timeseries added by vmagent for every scraped target.\n\nClick on the line and choose Drilldown to show CPU usage per instance\n", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [ + { + "title": "Drilldown", + "url": "/d/G7Z9GzMGz?viewPanel=123&var-job=${__field.labels.job}&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/out .*/" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 10 + }, + "id": 5, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate({__name__=~\"vmagent_rows_inserted_total|vm_promscrape_scraped_samples_sum\", job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job)", + "hide": false, + "interval": "", + "legendFormat": "in {{job}}", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmagent_remotewrite_block_size_rows_sum{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job)", + "interval": "", + "legendFormat": "out {{job}}", + "range": true, + "refId": "B" + } + ], + "title": "Samples rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the persistent queue size of pending samples in bytes >2MB which hasn't been flushed to remote storage yet. \n\nIncreasing of value might be a sign of connectivity issues. In such cases, vmagent starts to flush pending data on disk with attempt to send it later once connection is restored.\n\nRemote write URLs are hidden by default but might be unveiled once `-remoteWrite.showURL` is set to true.\n\nClick on the line and choose Drilldown to show the persistent queue size per instance.\n", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/G7Z9GzMGz?viewPanel=125&var-url=${__field.labels.url}&var-ds=$ds&var-instance=$instance&var-job=$job&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 10 + }, + "id": 17, + "links": [ + { + "title": "Troubleshooting", + "url": "https://docs.victoriametrics.com/vmagent/#troubleshooting" + } + ], + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(vmagent_remotewrite_pending_data_bytes{job=~\"$job\", instance=~\"$instance\", url=~\"$url\"}) by (job, url) > 2e6", + "interval": "", + "legendFormat": "{{job}} => {{url}}", + "range": true, + "refId": "A" + } + ], + "title": "Persistent queue size ($instance) to ($url)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of logging the messages by their level. Unexpected spike in rate is a good reason to check logs.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "bars", + "fillOpacity": 100, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 18 + }, + "id": 107, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_log_messages_total{job=~\"$job\",instance=~\"$instance\", level!=\"info\"}[$__rate_interval])) by (job, level) > 0", + "format": "time_series", + "hide": false, + "interval": "5m", + "intervalFactor": 1, + "legendFormat": "{{job}} - {{level}}", + "range": true, + "refId": "A" + } + ], + "title": "Logging rate", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows rate of dropped [samples](https://docs.victoriametrics.com/keyconcepts/#raw-samples) from persistent queue. vmagent drops samples from queue if in-memory and on-disk queues are full and it is unable to flush them to remote storage.\nThe max size of on-disk queue is configured by `-remoteWrite.maxDiskUsagePerURL` flag.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "decbytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 18 + }, + "id": 49, + "links": [ + { + "targetBlank": true, + "title": "Troubleshooting", + "url": "https://docs.victoriametrics.com/vmagent/#troubleshooting" + } + ], + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_persistentqueue_bytes_dropped_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (path) > 0", + "interval": "", + "legendFormat": "{{ path }}", + "range": true, + "refId": "A" + } + ], + "title": "Persistent queue dropped rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of requests served by vmagent HTTP server.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 26 + }, + "id": 15, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmagent_http_requests_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, path) > 0", + "interval": "", + "legendFormat": "", + "range": true, + "refId": "A" + } + ], + "title": "Requests rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Errors rate shows rate for multiple metrics that track possible errors in vmagent, such as network or parsing errors.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 26 + }, + "id": 69, + "links": [ + { + "targetBlank": true, + "title": "Troubleshooting", + "url": "https://docs.victoriametrics.com/vmagent/#troubleshooting" + } + ], + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmagent_http_request_errors_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, protocol) > 0", + "interval": "", + "legendFormat": "requests:{{protocol}} ({{job}})", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_protoparser_read_errors_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, type) > 0", + "interval": "", + "legendFormat": "parse: {{type}} ({{job}})", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_ingestserver_request_errors_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, type) > 0", + "interval": "", + "legendFormat": "ingest: {{type}} ({{job}})", + "range": true, + "refId": "C" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_protoparser_unmarshal_errors_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, type) > 0", + "interval": "", + "legendFormat": "unmarshal: {{type}} ({{job}})", + "range": true, + "refId": "D" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_promscrape_dial_errors_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "interval": "", + "legendFormat": "scrape dial ({{job}})", + "range": true, + "refId": "E" + } + ], + "title": "Errors rate ($instance)", + "type": "timeseries" + }, + { + "collapsed": true, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 34 + }, + "id": 45, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Click on the line and choose Drilldown to show CPU usage per instance", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/G7Z9GzMGz?viewPanel=119&var-job=${__field.labels.job}&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 3 + }, + "id": 109, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n rate(process_cpu_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_cpu_cores{job=~\"$job\", instance=~\"$instance\"}\n) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "CPU ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Percentage of used RSS memory (resident).\nThe RSS memory shows the amount of memory recently accessed by the application. It includes anonymous memory and data from recently accessed files (aka page cache).\nThe application's performance will significantly degrade when memory usage is close to 100%.\n\nClick on the line and choose Drilldown to show memory usage per instance", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/G7Z9GzMGz?viewPanel=117&var-job=${__field.labels.job}&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 3 + }, + "id": 111, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n max_over_time(process_resident_memory_bytes{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job\", instance=~\"$instance\"}\n) by(job)", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "RSS memory % usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of bytes read/write from the storage layer when vmagent has to buffer data on disk or read already buffered data.\n\nClick on the line and choose Drilldown to show CPU usage per instance", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [ + { + "targetBlank": true, + "title": "Drilldown", + "url": "/d/G7Z9GzMGz?viewPanel=121&var-job=${__field.labels.job}&var-ds=$ds&var-instance=$instance&${__url_time_range}" + } + ], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "read" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 11 + }, + "id": 81, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(process_io_storage_read_bytes_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "read {{job}}", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(process_io_storage_written_bytes_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "write {{job}}", + "range": true, + "refId": "B" + } + ], + "title": "Disk writes/reads ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Network usage shows the bytes rate for data accepted by vmagent and pushed via remotewrite protocol.\nDiscrepancies are possible because of different protocols used for ingesting, scraping and writing data.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bps" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/out.*/" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 11 + }, + "id": 7, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcplistener_read_bytes_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) * 8 \n+ sum(rate(vm_promscrape_conn_bytes_read_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) * 8", + "interval": "", + "legendFormat": "in {{job}}", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmagent_remotewrite_conn_bytes_written_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) * 8", + "interval": "", + "legendFormat": "out {{job}}", + "range": true, + "refId": "B" + } + ], + "title": "Network usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Panel shows the percentage of open file descriptors in the OS per instance.\nReaching the limit of open files (100%) can cause various issues and must be prevented.\n\nSee how to change limits here https://medium.com/@muhammadtriwibowo/set-permanently-ulimit-n-open-files-in-ubuntu-4d61064429a", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 5, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 19 + }, + "id": 83, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(\n max_over_time(process_open_fds{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n /\n process_max_fds{job=~\"$job\", instance=~\"$instance\"}\n) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Open FDs usage % ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 19 + }, + "id": 39, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(max_over_time(go_goroutines{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Goroutines ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the percent of CPU spent on garbage collection.\n\nIf % is high, then CPU usage can be decreased by changing GOGC to higher values. Increasing GOGC value will increase memory usage, and decrease CPU usage.\n\nTry searching for keyword `GOGC` at https://docs.victoriametrics.com/troubleshooting/ ", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 27 + }, + "id": 135, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(\n rate(go_gc_cpu_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]) \n / rate(process_cpu_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n ) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "CPU spent on GC ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the time goroutines have spent in runnable state before actually running. The lower is better.\n\nHigh values or values exceeding the threshold is usually a sign of insufficient CPU resources or CPU throttling. \n\nVerify that service has enough CPU resources. Otherwise, the service could work unreliably with delays in processing.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 0.1 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 27 + }, + "id": 149, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(histogram_quantile(0.99, sum(rate(go_sched_latencies_seconds_bucket{job=~\"$job\"}[$__rate_interval])) by (job, instance, le))) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Go scheduling latency", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 35 + }, + "id": 41, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(max_over_time(process_num_threads{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Threads ($instance)", + "type": "timeseries" + } + ], + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "refId": "A" + } + ], + "title": "Resource usage", + "type": "row" + }, + { + "collapsed": true, + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 35 + }, + "id": 94, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows top 10 jobs by the number of new series registered by vmagent over the 5min range. These jobs generate the most of the churn rate.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 4 + }, + "id": 92, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "topk_max(10, sum(sum_over_time(scrape_series_added[5m])) by (job)) > 0", + "interval": "", + "legendFormat": "{{ job }}", + "range": true, + "refId": "A" + } + ], + "title": "Top 10 jobs by unique samples", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows top 10 instances by the number of new series registered by vmagent over the 5min range. These instances generate the most of the churn rate.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 4 + }, + "id": 95, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "topk(10, sum(sum_over_time(scrape_series_added[5m])) by (instance)) > 0", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Top 10 instances by unique samples", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows write saturation of the persistent queue. If the threshold of 0.9sec is reached, then the persistent queue is saturated by more than 90% and vmagent won't be able to keep up with flushing data on disk. In this case, consider to decrease load on the vmagent or improve the disk throughput.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMax": 2, + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "transparent", + "value": null + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 12 + }, + "id": 98, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(rate(vm_persistentqueue_write_duration_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job)", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Persistent queue write saturation ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows read saturation of the persistent queue. If the threshold of 0.9sec is reached, then the persistent queue is saturated by more than 90% and vmagent won't be able to keep up with reading data from the disk. In this case, consider to decrease load on the vmagent or improve the disk throughput.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMax": 2, + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "transparent", + "value": null + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 12 + }, + "id": 99, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(rate(vm_persistentqueue_read_duration_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job)", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Persistent queue read saturation ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of dropped data blocks in cases when remote storage replies with `400 Bad Request` and `409 Conflict` HTTP responses.\n\nSee https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1149", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 20 + }, + "id": 79, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vmagent_remotewrite_packets_dropped_total{job=~\"$job\", instance=~\"$instance\", url=~\"$url\"}[$__rate_interval])) by(job, url) > 0", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Data blocks dropped ($instance) to ($url)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of dropped [samples](https://docs.victoriametrics.com/keyconcepts/#raw-samples) due to relabeling. \nMetric tracks drops for `-remoteWrite.relabelConfig` configuration only.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 20 + }, + "id": 18, + "links": [ + { + "targetBlank": true, + "title": "Relabeling", + "url": "https://docs.victoriametrics.com/vmagent/#relabeling" + } + ], + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmagent_remotewrite_relabel_metrics_dropped_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, url) > 0", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "B" + } + ], + "title": "Rows dropped by relabeling ($instance) to ($url)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of parsed datapoints from write or scrape requests.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 28 + }, + "id": 127, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_protoparser_rows_read_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, type) > 0", + "interval": "", + "legendFormat": "{{ type }} ({{job}})", + "range": true, + "refId": "A" + } + ], + "title": "Datapoints rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Tracks the rate of dropped invalid rows because of errors while unmarshaling write requests. The exact errors messages will be printed in logs.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 28 + }, + "id": 50, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_rows_invalid_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, type) > 0", + "interval": "", + "legendFormat": "{{type}} ({{job}})", + "range": true, + "refId": "A" + } + ], + "title": "Invalid datapoints rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "inspect": false + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Value" + }, + "properties": [ + { + "id": "custom.hidden", + "value": true + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Time" + }, + "properties": [ + { + "id": "custom.hidden", + "value": true + } + ] + } + ] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 36 + }, + "id": 129, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true, + "sortBy": [ + { + "desc": true, + "displayName": "job" + } + ] + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "sum(flag{is_set=\"true\", job=~\"$job\", instance=~\"$instance\"}) by(job, instance, name, value)", + "format": "table", + "instant": true, + "legendFormat": "__auto", + "range": false, + "refId": "A" + } + ], + "title": "Non-default flags", + "transformations": [ + { + "id": "groupBy", + "options": { + "fields": { + "instance": { + "aggregations": [] + }, + "job": { + "aggregations": [], + "operation": "groupby" + }, + "name": { + "aggregations": [], + "operation": "groupby" + }, + "value": { + "aggregations": [], + "operation": "groupby" + } + } + } + } + ], + "type": "table" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of restarts per job. The chart can be useful to identify periodic process restarts and correlate them with potential issues or anomalies. Normally, processes shouldn't restart unless restart was inited by user. The reason of restarts should be figured out by checking the logs of each specific service. ", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "stepAfter", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 36 + }, + "id": 150, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(changes(vm_app_start_timestamp{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]) > 0) by(job)", + "format": "time_series", + "instant": false, + "legendFormat": "{{job}}", + "refId": "A" + } + ], + "title": "Restarts ($job)", + "type": "timeseries" + } + ], + "targets": [ + { + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "refId": "A" + } + ], + "title": "Troubleshooting", + "type": "row" + }, + { + "collapsed": true, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 36 + }, + "id": 28, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 53 + }, + "id": 48, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(vm_promscrape_targets{job=~\"$job\", instance=~\"$instance\", status=\"up\"}) by(job, type) > 0", + "format": "time_series", + "interval": "", + "legendFormat": "{{type}} ({{job}})", + "range": true, + "refId": "A" + } + ], + "title": "Scrape targets UP(By Type)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 53 + }, + "id": 76, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(vm_promscrape_targets{job=~\"$job\", instance=~\"$instance\", status=\"down\"}) by(job, type) > 0", + "format": "time_series", + "interval": "", + "legendFormat": "{{type}} ({{job}})", + "range": true, + "refId": "A" + } + ], + "title": "Scrape targets DOWN(By Type)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 60 + }, + "id": 132, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(vm_promscrape_scrape_pool_targets{job=~\"$job\", instance=~\"$instance\", status=\"up\"}) by(job, scrape_job) > 0", + "format": "time_series", + "interval": "", + "legendFormat": "{{job}}: {{scrape_job}}", + "range": true, + "refId": "A" + } + ], + "title": "Scrape targets UP(By Job)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 60 + }, + "id": 133, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(vm_promscrape_scrape_pool_targets{job=~\"$job\", instance=~\"$instance\", status=\"down\"}) by(job, scrape_job) > 0", + "format": "time_series", + "interval": "", + "legendFormat": "{{job}}: {{scrape_job}}", + "range": true, + "refId": "A" + } + ], + "title": "Scrape targets DOWN(By Job)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of scrapes per second.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 67 + }, + "id": 20, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_promscrape_scrapes_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job)", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Scrape rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of datapoints scraped per second.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 67 + }, + "id": 126, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_promscrape_scraped_samples_sum{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job)", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Scraped datapoints rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 75 + }, + "id": 46, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(histogram_quantile(0.99, sum(rate(vm_promscrape_scrape_response_size_bytes_bucket{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, vmrange))) by(job)", + "format": "time_series", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Scrape response size 0.99 quantile ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 75 + }, + "id": 148, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(histogram_quantile(0.99, sum(rate(vm_promscrape_scrape_duration_seconds_bucket{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, vmrange))) by(job)", + "format": "time_series", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Scrape duration 0.99 quantile ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 83 + }, + "id": 31, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_promscrape_scrapes_failed_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "interval": "", + "legendFormat": "scrapes failed ({{job}})", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_promscrape_scrapes_timed_out_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "interval": "", + "legendFormat": "timeouts ({{job}})", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_promscrape_scrapes_gunzip_failed_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "interval": "", + "legendFormat": "gunzip fails ({{job}})", + "range": true, + "refId": "C" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_promscrape_dial_errors_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "interval": "", + "legendFormat": "dial fails ({{job}})", + "range": true, + "refId": "D" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_promscrape_max_scrape_size_exceeded_errors_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "hide": false, + "interval": "", + "legendFormat": "max scrape size exceeded ({{job}})", + "range": true, + "refId": "E" + } + ], + "title": "Scrape fails ($instance)", + "type": "timeseries" + } + ], + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "refId": "A" + } + ], + "title": "Scraping", + "type": "row" + }, + { + "collapsed": true, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 37 + }, + "id": 71, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of write requests served by ingestserver (UDP, TCP connections) and HTTP server.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 30 + }, + "id": 73, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vm_ingestserver_requests_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, type, net) > 0", + "interval": "", + "legendFormat": "{{net}}: {{ type }} ({{job}})", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vmagent_http_requests_total{job=~\"$job\", instance=~\"$instance\", protocol!=\"\"}[$__rate_interval])) by(job, protocol) > 0", + "interval": "", + "legendFormat": "{{ protocol }}: http ({{job}})", + "range": true, + "refId": "B" + } + ], + "title": "Requests rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of rows ingested in vmagent via push protocols.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 22 + }, + "id": 131, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vmagent_rows_inserted_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, type) > 0", + "interval": "", + "legendFormat": "{{ type }} ({{job}})", + "range": true, + "refId": "A" + } + ], + "title": "Rows rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows how many concurrent inserts are taking place.\n\nIf the number of concurrent inserts hitting the `limit` or is close to the `limit` constantly - it might be a sign of a resource shortage.\n\n If vmagent's CPU usage and remote write connection saturation are at normal level, it might be that `-maxConcurrentInserts` cmd-line flag need to be increased.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 30 + }, + "id": 130, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max_over_time(vm_concurrent_insert_current{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])", + "interval": "", + "legendFormat": "{{instance}} ({{job}})", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "min(vm_concurrent_insert_capacity{job=~\"$job\", instance=~\"$instance\"}) by(job)", + "interval": "", + "legendFormat": "limit ({{job}})", + "range": true, + "refId": "B" + } + ], + "title": "Concurrent inserts ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of write errors in ingestserver (UDP, TCP connections) and HTTP server.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 30 + }, + "id": 77, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": true, + "expr": "sum(rate(vm_ingestserver_request_errors_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(type, net) > 0", + "interval": "", + "legendFormat": "{{ type }} ({{net}})", + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": true, + "expr": "sum(rate(vmagent_http_request_errors_total{job=~\"$job\", instance=~\"$instance\", protocol!=\"\"}[$__rate_interval])) by(protocol) > 0", + "interval": "", + "legendFormat": "{{ protocol }} (http)", + "refId": "B" + } + ], + "title": "Error rate ($instance)", + "type": "timeseries" + } + ], + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "refId": "A" + } + ], + "title": "Ingestion", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 38 + }, + "id": 136, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of matched samples by the aggregation rule. \n\nThe more samples is matched, the more work this aggregation rule does. The matching rule is specified via `match` param.\n\nSee more details in [stream aggregation config](https://docs.victoriametrics.com/stream-aggregation/#stream-aggregation-config). ", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 23 + }, + "id": 146, + "options": { + "legend": { + "calcs": ["min", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_streamaggr_matched_samples_total{job=~\"$job\",instance=~\"$instance\", url=~\"$url\"}[$__rate_interval])) without (instance, pod) > 0", + "instant": false, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Matched samples ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "The rate of ignored samples during aggregation. \nStream aggregation will drop samples with NaN values, or samples with too old timestamps. See https://docs.victoriametrics.com/stream-aggregation/#ignoring-old-samples ", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 23 + }, + "id": 143, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_streamaggr_ignored_samples_total{job=~\"$job\",instance=~\"$instance\", url=~\"$url\"}[$__rate_interval]) > 0) without (instance, pod)", + "instant": false, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Ignored samples ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of produced samples by the aggregation rule. \n\nNumber of produced samples depend on params like `by`, `without`, `interval`, etc.\n\nSee more details in [stream aggregation config](https://docs.victoriametrics.com/stream-aggregation/#stream-aggregation-config). ", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 31 + }, + "id": 147, + "options": { + "legend": { + "calcs": ["min", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_streamaggr_output_samples_total{job=~\"$job\",instance=~\"$instance\", url=~\"$url\"}[$__rate_interval])) without (instance, pod) > 0", + "instant": false, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Produced samples ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows events when deduplication or aggregation couldn't be finished in the configured interval. Such events may result into bad accuracy of produced data.\n\nPossible solutions:\n* increase interval; \n* use match filter matching smaller number of series;\n* reduce samples ingestion rate to stream aggregation", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "min": -5, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 31 + }, + "id": 139, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "increase(vm_streamaggr_flush_timeouts_total{job=~\"$job\",instance=~\"$instance\", url=~\"$url\"}[$__rate_interval]) > 0", + "instant": false, + "legendFormat": "aggregation: {{url}} ({{job}}): {{path}}:{{position}}", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "increase(vm_streamaggr_dedup_flush_timeouts_total{job=~\"$job\",instance=~\"$instance\", url=~\"$url\"}[$__rate_interval]) > 0", + "hide": false, + "instant": false, + "legendFormat": "deduplication: {{url}} ({{job}}): {{path}}:{{position}}", + "range": true, + "refId": "B" + } + ], + "title": "Flush timeouts ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the max lag between samples timestamps within one batch passed to the aggregation input.\n\nLower is better.\n\nToo high lag or lag exceeding the interval might be a sign that data was delayed before aggregation or resource insufficiency on aggregator. Samples with high lag may affect accuracy of aggregation.\n\nSee https://docs.victoriametrics.com/stream-aggregation/#ignoring-old-samples", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 39 + }, + "id": 142, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "histogram_quantile(0.99, sum(rate(vm_streamaggr_samples_lag_seconds_bucket{job=~\"$job\",instance=~\"$instance\", url=~\"$url\"}[$__rate_interval])) without (instance, pod))", + "instant": false, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Samples lag 0.99 quantile ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "The 99th percentile of avg flush duration for the aggregated data. \n\nSmaller is better.\n\nAggregation can produce incorrect results ff flush duration exceeds configured deduplication interval. See \"Flush Timeouts\" panel.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 39 + }, + "id": 137, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "histogram_quantile(0.99, rate(vm_streamaggr_dedup_flush_duration_seconds_bucket{job=~\"$job\",instance=~\"$instance\", url=~\"$url\"}[$__rate_interval]))", + "instant": false, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Dedup flush duration 0.99 quantile ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the size of Label Compressor in number of entries.\n\nLabels compressor encodes label-value pairs during aggregation to optimise memory usage. It is expected for its size to grow with time and to reset on vmagent restarts.\n\nRapid spikes in Label compressor size might be a sign of significant changes in labels of received samples.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/bytes.*/" + }, + "properties": [ + { + "id": "custom.axisPlacement", + "value": "right" + }, + { + "id": "unit", + "value": "bytes" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 47 + }, + "id": 141, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(vm_streamaggr_labels_compressor_items_count{job=~\"$job\",instance=~\"$instance\"}) by(job, instance)", + "hide": false, + "instant": false, + "legendFormat": "items: {{instance}} ({{job}})", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(vm_streamaggr_labels_compressor_size_bytes{job=~\"$job\", instance=~\"$instance\"}) by(job, instance)", + "hide": false, + "instant": false, + "legendFormat": "bytes: {{instance}} ({{job}})", + "range": true, + "refId": "B" + } + ], + "title": "Labels compressor ($instance)", + "type": "timeseries" + } + ], + "title": "Streaming aggregation", + "type": "row" + }, + { + "collapsed": true, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 39 + }, + "id": 58, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of requests to configured remote write endpoints by url and status code.\n\nRemote write URLs are hidden by default but might be unveiled once `-remoteWrite.showURL` is set to true.\n\n", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 2, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 24 + }, + "id": 60, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vmagent_remotewrite_requests_total{job=~\"$job\", instance=~\"$instance\", url=~\"$url\"}[$__rate_interval])) by(job, url, status_code) > 0", + "interval": "", + "legendFormat": "", + "range": true, + "refId": "A" + } + ], + "title": "Requests rate ($instance) to ($url)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the global rate for number of written bytes via remote write connections.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "decbytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 24 + }, + "id": 66, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vmagent_remotewrite_conn_bytes_written_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job) > 0", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Bytes write rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows requests retry rate by url. Number of retries is unlimited but protected with delays up to 1m between attempts.\n\nRemote write URLs are hidden by default but might be unveiled once `-remoteWrite.showURL` is set to true.\n\n", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 32 + }, + "id": 61, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(rate(vmagent_remotewrite_retries_count_total{job=~\"$job\", instance=~\"$instance\", url=~\"$url\"}[$__rate_interval])) by(url) > 0", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Retry rate ($instance) to ($url)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows current number of established connections to remote write endpoints.\n\n", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 32 + }, + "id": 65, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(max_over_time(vmagent_remotewrite_conns{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job)", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Connections ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the current limit usage of unique series over an hourly period. Vmagent will start to drop series once the limit is reached.\n\nPlease note, panel will be blank if `remoteWrite.maxHourlySeries` is not set.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "mappings": [], + "max": 100, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "transparent" + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 40 + }, + "id": 88, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n vmagent_hourly_series_limit_current_series{job=~\"$job\", instance=~\"$instance\"} \n / \n vmagent_hourly_series_limit_max_series{job=~\"$job\", instance=~\"$instance\"}\n ) by(job) * 100", + "interval": "", + "legendFormat": "{{job}}", + "range": true, + "refId": "A" + } + ], + "title": "Hourly series limit", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows saturation of every connection to remote storage. If the threshold of 90% is reached, then the connection is saturated (busy or slow) by more than 90%, so vmagent won't be able to keep up and can start buffering data. \n\nThis usually means that `-remoteWrite.queues` command-line flag must be increased in order to increase the number of connections per each remote storage.\n", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "transparent" + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 40 + }, + "id": 84, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n rate(vmagent_remotewrite_send_duration_seconds_total{job=~\"$job\", instance=~\"$instance\", url=~\"$url\"}[$__rate_interval])\n /\n vmagent_remotewrite_queues{job=~\"$job\", instance=~\"$instance\", url=~\"$url\"}\n) by(job, url)", + "interval": "", + "legendFormat": "", + "range": true, + "refId": "A" + } + ], + "title": "Remote write connection saturation ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the current limit usage of unique series over a daily period. Vmagent will start to drop series once the limit is reached.\n\nPlease note, panel will be blank if `remoteWrite.maxDailySeries` is not set.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "line" + } + }, + "mappings": [], + "max": 100, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "transparent" + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 48 + }, + "id": 90, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max(\n vmagent_daily_series_limit_current_series{job=~\"$job\",instance=~\"$instance\"} \n / \n vmagent_daily_series_limit_max_series{job=~\"$job\",instance=~\"$instance\"}\n) by(job)", + "interval": "", + "legendFormat": "{{job}}", + "range": true, + "refId": "A" + } + ], + "title": "Daily series limit", + "type": "timeseries" + } + ], + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "refId": "A" + } + ], + "title": "Remote write", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 40 + }, + "id": 113, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "gridPos": { + "h": 2, + "w": 24, + "x": 0, + "y": 33 + }, + "id": 115, + "options": { + "code": { + "language": "plaintext", + "showLineNumbers": false, + "showMiniMap": false + }, + "content": "Drilldown row is used by other panels on the dashboard to show more detailed metrics per-instance.", + "mode": "markdown" + }, + "pluginVersion": "10.4.2", + "transparent": true, + "type": "text" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 35 + }, + "id": 119, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "sum(rate(process_cpu_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{instance}} ({{job}})", + "range": true, + "refId": "A" + } + ], + "title": "CPU usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the used memory (resident).\nThe application's performance will significantly degrade when memory usage is close to 100%.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + } + ] + }, + "unit": "decbytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 35 + }, + "id": 117, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "max_over_time(process_resident_memory_bytes{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])", + "interval": "", + "legendFormat": "{{instance}} ({{job}})", + "range": true, + "refId": "A" + } + ], + "title": "RSS memory usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the persistent queue size of pending samples in bytes which hasn't been flushed to remote storage yet. \n\nIncreasing of value might be a sign of connectivity issues. In such cases, vmagent starts to flush pending data on disk with attempt to send it later once connection is restored.\n\nRemote write URLs are hidden by default but might be unveiled once `-remoteWrite.showURL` is set to true.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 43 + }, + "id": 125, + "links": [ + { + "title": "Troubleshooting", + "url": "https://docs.victoriametrics.com/vmagent/#troubleshooting" + } + ], + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": true, + "expr": "sum(vmagent_remotewrite_pending_data_bytes{job=~\"$job\", instance=~\"$instance\", url=~\"$url\"}) by (instance, url)", + "interval": "", + "legendFormat": "{{instance}} => {{url}}", + "range": true, + "refId": "A" + } + ], + "title": "Persistent queue size ($instance) to ($url)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows in/out samples rate including push and pull models. \n\nThe out-rate could be different to in-rate because of replication or additional timeseries added by vmagent for every scraped target.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/out .*/" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 43 + }, + "id": 123, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_promscrape_scraped_samples_sum{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)\n+ sum(rate(vmagent_rows_inserted_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)", + "hide": false, + "interval": "", + "legendFormat": "in {{instance}} {{job}}", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmagent_remotewrite_block_size_rows_sum{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)", + "interval": "", + "legendFormat": "out {{instance}} {{job}}", + "range": true, + "refId": "B" + } + ], + "title": "Samples rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the number of bytes read/write from the storage layer when vmagent has to buffer data on disk or read already buffered data.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "read" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 51 + }, + "id": 121, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(process_io_storage_read_bytes_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "read {{instance}} {{job}}", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(process_io_storage_written_bytes_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job,instance)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "write {{instance}} {{job}}", + "range": true, + "refId": "B" + } + ], + "title": "Disk writes/reads ($instance)", + "type": "timeseries" + } + ], + "title": "Drilldown", + "type": "row" + } + ], + "refresh": "", + "schemaVersion": 39, + "tags": ["vmagent", "victoriametrics"], + "templating": { + "list": [ + { + "current": { + "selected": false, + "text": "VictoriaMetrics - cluster", + "value": "PAF93674D0B4E9963" + }, + "hide": 0, + "includeAll": false, + "multi": false, + "name": "ds", + "options": [], + "query": "prometheus", + "queryValue": "", + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "type": "datasource" + }, + { + "allValue": "", + "current": {}, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "definition": "label_values(vm_app_version{version=~\"^vmagent.*\"}, job)", + "hide": 0, + "includeAll": true, + "multi": true, + "name": "job", + "options": [], + "query": { + "query": "label_values(vm_app_version{version=~\"^vmagent.*\"}, job)", + "refId": "VictoriaMetrics-job-Variable-Query" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "tagValuesQuery": "", + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": ".*", + "current": {}, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "definition": "label_values(vm_app_version{job=~\"$job\"}, instance)", + "hide": 0, + "includeAll": true, + "multi": true, + "name": "instance", + "options": [], + "query": { + "query": "label_values(vm_app_version{job=~\"$job\"}, instance)", + "refId": "VictoriaMetrics-instance-Variable-Query" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "tagValuesQuery": "", + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": ".*", + "current": {}, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "definition": "label_values(vmagent_remotewrite_requests_total{job=~\"$job\", instance=~\"$instance\"}, url)", + "description": "The remote write URLs", + "hide": 0, + "includeAll": true, + "multi": true, + "name": "url", + "options": [], + "query": { + "query": "label_values(vmagent_remotewrite_requests_total{job=~\"$job\", instance=~\"$instance\"}, url)", + "refId": "StandardVariableQuery" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "type": "query" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "filters": [], + "hide": 0, + "name": "adhoc", + "skipUrlSync": false, + "type": "adhoc" + } + ] + }, + "time": { + "from": "now-3h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ] + }, + "timezone": "", + "title": "VictoriaMetrics - vmagent", + "uid": "G7Z9GzMGz", + "version": 1, + "weekStart": "", + "gnetId": 12683 +} diff --git a/rootfs/usr/share/grafana/dashboards/drycc_victoriametrics_vmauth.json b/rootfs/usr/share/grafana/dashboards/drycc_victoriametrics_vmauth.json new file mode 100644 index 0000000..f26e534 --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/drycc_victoriametrics_vmauth.json @@ -0,0 +1,2576 @@ +{ + "__inputs": [], + "__elements": {}, + "__requires": [ + { + "type": "grafana", + "id": "grafana", + "name": "Grafana", + "version": "10.4.2" + }, + { + "type": "datasource", + "id": "prometheus", + "name": "Prometheus", + "version": "1.0.0" + }, + { + "type": "panel", + "id": "stat", + "name": "Stat", + "version": "" + }, + { + "type": "panel", + "id": "table", + "name": "Table", + "version": "" + }, + { + "type": "panel", + "id": "timeseries", + "name": "Time series", + "version": "" + } + ], + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": { + "type": "grafana", + "uid": "-- Grafana --" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "target": { + "limit": 100, + "matchAny": false, + "tags": [], + "type": "dashboard" + }, + "type": "dashboard" + } + ] + }, + "description": "Overview for VictoriaMetrics vmauth v1.80.0 or higher", + "editable": false, + "fiscalYearStartMonth": 0, + "graphTooltip": 1, + "id": null, + "links": [ + { + "asDropdown": false, + "icon": "doc", + "includeVars": false, + "keepTime": false, + "tags": [], + "targetBlank": true, + "title": "vmauth docs", + "tooltip": "vmauth docs", + "type": "link", + "url": "https://docs.victoriametrics.com/vmauth.html" + }, + { + "asDropdown": false, + "icon": "question", + "includeVars": false, + "keepTime": false, + "tags": [], + "targetBlank": true, + "title": "Found a bug?", + "tooltip": "Found a bug?", + "type": "link", + "url": "https://github.com/VictoriaMetrics/VictoriaMetrics/issues" + }, + { + "asDropdown": false, + "icon": "info", + "includeVars": false, + "keepTime": false, + "tags": [], + "targetBlank": true, + "title": "New releases", + "tooltip": "New releases", + "type": "link", + "url": "https://github.com/VictoriaMetrics/VictoriaMetrics/releases" + } + ], + "liveNow": false, + "panels": [ + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 2, + "panels": [], + "title": "Stats", + "type": "row" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "stepAfter", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 11, + "x": 0, + "y": 1 + }, + "id": 32, + "options": { + "legend": { + "calcs": [ + "lastNotNull" + ], + "displayMode": "table", + "placement": "right", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "asc" + } + }, + "pluginVersion": "9.2.6", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "sum(min_over_time(vm_app_uptime_seconds{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job)", + "format": "time_series", + "instant": false, + "interval": "", + "legendFormat": "{{job}}", + "refId": "A" + } + ], + "title": "Uptime", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows if the last configuration update was successful. \"Not Ok\" means there was an unsuccessful attempt to update the configuration due to some error. Check the log for details.", + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "0": { + "color": "green", + "index": 0, + "text": "Ok" + } + }, + "type": "value" + }, + { + "options": { + "from": 1, + "result": { + "color": "red", + "index": 1, + "text": "Not Ok" + }, + "to": 999999 + }, + "type": "range" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 11, + "y": 1 + }, + "id": 30, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "last" + ], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": false, + "expr": "count(vmauth_config_last_reload_successful{job=~\"$job\", instance=~\"$instance\"} < 1 ) or 0", + "interval": "", + "legendFormat": "", + "refId": "A" + } + ], + "title": "Config update", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of requests.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 7, + "x": 17, + "y": 1 + }, + "id": 4, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "(sum(rate(vmauth_user_requests_total{job=~\"$job\", instance=~\"$instance\", username=~\"$user\"}[$__rate_interval])) or 0) + (sum(rate(vmauth_unauthorized_user_requests_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) or 0)", + "instant": true, + "range": false, + "refId": "A" + } + ], + "title": "Requests rate", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the total number of users defined at configuration file.", + "fieldConfig": { + "defaults": { + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 11, + "y": 4 + }, + "id": 31, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "last" + ], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "count(vmauth_user_concurrent_requests_capacity{job=~\"$job\", instance=~\"$instance\"})", + "interval": "", + "legendFormat": "", + "range": true, + "refId": "A" + } + ], + "title": "Users count", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of request errors.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 7, + "x": 17, + "y": 4 + }, + "id": 36, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "sum(rate(vmauth_http_request_errors_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]))", + "instant": true, + "range": false, + "refId": "A" + } + ], + "title": "Errors rate", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "filterable": false, + "inspect": false, + "minWidth": 50 + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Time" + }, + "properties": [ + { + "id": "custom.hidden", + "value": true + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Value" + }, + "properties": [ + { + "id": "displayName", + "value": "Count" + } + ] + } + ] + }, + "gridPos": { + "h": 4, + "w": 11, + "x": 0, + "y": 5 + }, + "id": 6, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": [ + "sum" + ], + "show": false + }, + "frameIndex": 0, + "showHeader": true + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "sum(vm_app_version{job=~\"$job\", instance=~\"$instance\"}) by (job, short_version)", + "format": "table", + "instant": true, + "range": false, + "refId": "A" + } + ], + "title": "Version", + "type": "table" + }, + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 9 + }, + "id": 13, + "panels": [], + "title": "Overview", + "type": "row" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 10 + }, + "id": 11, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmauth_user_requests_total{job=~\"$job\", instance=~\"$instance\", username=~\"$user\"}[$__rate_interval])) by(username)", + "hide": false, + "legendFormat": "__auto", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmauth_unauthorized_user_requests_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]))", + "hide": false, + "legendFormat": "__auto", + "range": true, + "refId": "B" + } + ], + "title": "Requests rate", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows percent utilization of per concurrent requests capacity.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "bars", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "dashed" + } + }, + "mappings": [], + "max": 1, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 10 + }, + "id": 14, + "options": { + "legend": { + "calcs": [ + "max", + "mean" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Mean", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(\nmax_over_time(vmauth_user_concurrent_requests_current{job=~\"$job\", instance=~\"$instance\", username=~\"$user\"}[$__rate_interval])\n/ \nvmauth_user_concurrent_requests_capacity{job=~\"$job\", instance=~\"$instance\", username=~\"$user\"}\n) by(username) > 0\n", + "hide": false, + "interval": "5m", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "User concurrent requests usage", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the rate of rejected requests by a reason.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 19 + }, + "id": 16, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmauth_http_request_errors_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (reason)", + "hide": false, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Requests rejected rate", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": " The number of concurrent connections processed by vmauth reached one of limits. Possible solutions:\n- increase global limit with flag -maxConcurrentRequests\n- increase limit with flag: -maxConcurrentPerUserRequests for all users or with config option `max_concurrent_requests` per user.\n- deploy additional vmauth replicas\n- check requests latency at backend service and allocate resources to it if needed", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "bars", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 19 + }, + "id": 10, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmauth_user_concurrent_requests_limit_reached_total{job=~\"$job\", instance=~\"$instance\", username=~\"$user\"}[$__rate_interval])) by(username) > 0", + "interval": "1m", + "legendFormat": "__auto", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmauth_unauthorized_user_concurrent_requests_limit_reached_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) > 0", + "hide": false, + "legendFormat": "unauthorized", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vmauth_concurrent_requests_limit_reached_total[$__rate_interval])) > 0", + "hide": false, + "legendFormat": "global at {{ $instance }}", + "range": true, + "refId": "C" + } + ], + "title": "Concurrent limit reached", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows duration in seconds of user requests by quantile.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 28 + }, + "id": 19, + "options": { + "legend": { + "calcs": [ + "max", + "mean" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Max", + "sortDesc": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(vmauth_user_request_duration_seconds{job=~\"$job\", instance=~\"$instance\", username=~\"$user\", quantile=~\"(0.99|0.5)\"}) by (quantile, username) > 0", + "hide": false, + "legendFormat": "user: {{username}} q: {{ quantile}}", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max(vmauth_unauthorized_user_request_duration_seconds{job=~\"$job\", instance=~\"$instance\", quantile=~\"(0.99|0.5)\"}) by (quantile) > 0", + "hide": false, + "legendFormat": "user: unauthorized q: {{ quantile}}", + "range": true, + "refId": "B" + } + ], + "title": "User requests duration", + "type": "timeseries" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 37 + }, + "id": 22, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Percentage of used memory (resident).\nThe application's performance will significantly degrade when memory usage is close to 100%.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 11 + }, + "id": 25, + "options": { + "legend": { + "calcs": [ + "mean", + "lastNotNull", + "max" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "max(\n max_over_time(process_resident_memory_bytes{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_memory_bytes{job=~\"$job\", instance=~\"$instance\"}\n) by(job)", + "interval": "", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "RSS memory % usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 11 + }, + "id": 26, + "options": { + "legend": { + "calcs": [ + "mean", + "lastNotNull", + "max" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "max(\n rate(process_cpu_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n /\n vm_available_cpu_cores{job=~\"$job\", instance=~\"$instance\"}\n) by(job)", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "CPU % usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 19 + }, + "id": 27, + "options": { + "legend": { + "calcs": [ + "mean", + "lastNotNull", + "max" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(go_memstats_sys_bytes{job=~\"$job\", instance=~\"$instance\"}) + sum(vm_cache_size_bytes{job=~\"$job\", instance=~\"$instance\"})", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "requested from system", + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(go_memstats_heap_inuse_bytes{job=~\"$job\", instance=~\"$instance\"}) + sum(vm_cache_size_bytes{job=~\"$job\", instance=~\"$instance\"})", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "heap inuse", + "refId": "B" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(go_memstats_stack_inuse_bytes{job=~\"$job\", instance=~\"$instance\"})", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "stack inuse", + "refId": "C" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(process_resident_memory_bytes{job=~\"$job\", instance=~\"$instance\"})", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "resident", + "refId": "D" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": false, + "expr": "sum(process_resident_memory_anon_bytes{job=~\"$job\", instance=~\"$instance\"})", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "resident anonymous", + "refId": "E" + } + ], + "title": "Memory usage ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Limit" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#F2495C", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 19 + }, + "id": 28, + "options": { + "legend": { + "calcs": [ + "mean", + "lastNotNull", + "max" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "rate(process_cpu_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "CPU cores used", + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "exemplar": false, + "expr": "process_cpu_cores_available{job=~\"$job\", instance=~\"$instance\"}", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "Limit", + "refId": "B" + } + ], + "title": "CPU ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 27 + }, + "id": 23, + "options": { + "legend": { + "calcs": [ + "mean", + "lastNotNull", + "max" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(vm_tcplistener_conns{job=~\"$job\", instance=~\"$instance\"})", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "connections", + "refId": "A" + } + ], + "title": "TCP connections ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 27 + }, + "id": 24, + "options": { + "legend": { + "calcs": [ + "mean", + "lastNotNull", + "max" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(rate(vm_tcplistener_accepts_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "connections", + "range": true, + "refId": "A" + } + ], + "title": "TCP connections rate ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows the percentage of open file descriptors compared to the limit set in the OS.\nReaching the limit of open files can cause various issues and must be prevented.\n\nSee how to change limits here https://medium.com/@muhammadtriwibowo/set-permanently-ulimit-n-open-files-in-ubuntu-4d61064429a", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 2, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "max" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#C4162A", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 35 + }, + "id": 20, + "options": { + "legend": { + "calcs": [ + "mean", + "lastNotNull", + "max" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "max_over_time(process_open_fds{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])\n/\nprocess_max_fds{job=~\"$job\", instance=~\"$instance\"}", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{job}}", + "range": true, + "refId": "A" + } + ], + "title": "Open FDs ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 35 + }, + "id": 21, + "options": { + "legend": { + "calcs": [ + "mean", + "lastNotNull", + "max" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(go_goroutines{job=~\"$job\", instance=~\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "gc duration", + "refId": "A" + } + ], + "title": "Goroutines ($instance)", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 43 + }, + "id": 29, + "options": { + "legend": { + "calcs": [ + "mean", + "lastNotNull", + "max" + ], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "9.1.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "expr": "sum(process_num_threads{job=~\"$job\", instance=~\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "threads", + "refId": "A" + } + ], + "title": "Threads ($instance)", + "type": "timeseries" + } + ], + "title": "Resource usage", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 38 + }, + "id": 35, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "inspect": false + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Value" + }, + "properties": [ + { + "id": "custom.hidden", + "value": true + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Time" + }, + "properties": [ + { + "id": "custom.hidden", + "value": true + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 39 + }, + "id": 34, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": [ + "sum" + ], + "show": false + }, + "showHeader": true, + "sortBy": [ + { + "desc": true, + "displayName": "job" + } + ] + }, + "pluginVersion": "10.4.2", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "exemplar": false, + "expr": "sum(flag{is_set=\"true\", job=~\"$job\", instance=~\"$instance\"}) by(job, instance, name, value)", + "format": "table", + "instant": true, + "legendFormat": "__auto", + "range": false, + "refId": "A" + } + ], + "title": "Non-default flags", + "transformations": [ + { + "id": "groupBy", + "options": { + "fields": { + "instance": { + "aggregations": [] + }, + "job": { + "aggregations": [] + }, + "name": { + "aggregations": [], + "operation": "groupby" + }, + "value": { + "aggregations": [], + "operation": "groupby" + } + } + } + } + ], + "type": "table" + }, + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "description": "Shows number of generated error and warning messages in logs. Non-zero value may be a sign of connectivity or missconfiguration errors.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 39 + }, + "id": 8, + "options": { + "legend": { + "calcs": [ + "max", + "last", + "mean" + ], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "10.4.0", + "targets": [ + { + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "editorMode": "code", + "expr": "sum(increase(vm_log_messages_total{job=~\"$job\", instance=~\"$instance\", level!=\"info\"}[$__rate_interval])) by (level)", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Log errors", + "type": "timeseries" + } + ], + "title": "Troubleshooting", + "type": "row" + } + ], + "refresh": "30s", + "revision": 1, + "schemaVersion": 39, + "tags": [], + "templating": { + "list": [ + { + "current": { + "selected": false, + "text": "VictoriaMetrics - cluster", + "value": "PAF93674D0B4E9963" + }, + "hide": 0, + "includeAll": false, + "multi": false, + "name": "ds", + "options": [], + "query": "prometheus", + "queryValue": "", + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "type": "datasource" + }, + { + "current": {}, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "definition": "label_values(vm_app_version{version=~\"^vmauth.*\"}, job)", + "hide": 0, + "includeAll": true, + "multi": true, + "name": "job", + "options": [], + "query": { + "query": "label_values(vm_app_version{version=~\"^vmauth.*\"}, job)", + "refId": "StandardVariableQuery" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "type": "query" + }, + { + "allValue": ".*", + "current": {}, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "definition": "label_values(vm_app_version{job=~\"$job\"}, instance)", + "hide": 0, + "includeAll": true, + "multi": true, + "name": "instance", + "options": [], + "query": { + "query": "label_values(vm_app_version{job=~\"$job\"}, instance)", + "refId": "StandardVariableQuery" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "type": "query" + }, + { + "current": {}, + "datasource": { + "type": "prometheus", + "uid": "$ds" + }, + "definition": "label_values(vmauth_user_requests_total{job=~\"$job\", instance=~\"$instance\"}, username)", + "hide": 0, + "includeAll": true, + "multi": true, + "name": "user", + "options": [], + "query": { + "query": "label_values(vmauth_user_requests_total{job=~\"$job\", instance=~\"$instance\"}, username)", + "refId": "StandardVariableQuery" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "type": "query" + }, + { + "datasource": { + "type": "victoriametrics-datasource", + "uid": "P38648FE0F8C5BEA2" + }, + "filters": [], + "hide": 0, + "name": "adhoc", + "skipUrlSync": false, + "type": "adhoc" + } + ] + }, + "time": { + "from": "now-3h", + "to": "now" + }, + "timepicker": {}, + "timezone": "", + "title": "VictoriaMetrics - vmauth", + "uid": "nbuo5Mr4k", + "version": 1, + "weekStart": "", + "gnetId": 21394 +} \ No newline at end of file diff --git a/rootfs/usr/share/grafana/dashboards/kubernetes.json b/rootfs/usr/share/grafana/dashboards/kubernetes.json new file mode 100644 index 0000000..af96df7 --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/kubernetes.json @@ -0,0 +1,5926 @@ +{ + "annotations": { + "list": [ + { + "$$hashKey": "object:247", + "builtIn": 1, + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "target": { + "limit": 100, + "matchAny": false, + "tags": [], + "type": "dashboard" + }, + "type": "dashboard" + } + ] + }, + "description": "【English version】2025.01.25 Updates，kubernetesComprehensive display of resources！Includes K8S Overall Resource Overview、Microservices Resource Details、Pod Resource Details andK8SNetwork Bandwidth，Optimization metrics。https://grafana.com/orgs/starsliao/dashboards", + "editable": false, + "fiscalYearStartMonth": 0, + "graphTooltip": 0, + "id": null, + "links": [ + { + "icon": "bolt", + "tags": [], + "targetBlank": true, + "title": "Update", + "tooltip": "See more dashboards", + "type": "link", + "url": "https://grafana.com/orgs/starsliao/dashboards" + }, + { + "$$hashKey": "object:831", + "icon": "question", + "tags": ["node_exporter"], + "targetBlank": true, + "title": "GitHub", + "tooltip": "View more dashboards", + "type": "link", + "url": "https://github.com/starsliao" + } + ], + "panels": [ + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 54, + "panels": [], + "title": "Node Resource Overview：Selected Nodes:【$Node】", + "type": "row" + }, + { + "fieldConfig": { + "defaults": { + "decimals": 1, + "mappings": [], + "max": 1, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "orange", + "value": 0.8 + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 4, + "x": 0, + "y": 1 + }, + "id": 44, + "options": { + "displayMode": "basic", + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "maxVizHeight": 300, + "minVizHeight": 10, + "minVizWidth": 0, + "namePlacement": "auto", + "orientation": "vertical", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showUnfilled": false, + "sizing": "auto", + "text": {}, + "valueMode": "color" + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",container!=\"\",node=~\"^$Node$\"}) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\", node=~\"^$Node$\"})", + "format": "time_series", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Memory Utilization", + "refId": "C" + }, + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",node=~\"^$Node$\"}) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\", node=~\"^$Node$\"})", + "format": "time_series", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Memory Request Rate", + "refId": "A" + }, + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",node=~\"^$Node$\"}) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\", node=~\"^$Node$\"})", + "format": "time_series", + "hide": false, + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "Memory Limit Rate", + "refId": "B", + "step": 10 + } + ], + "title": "Node Memory Ratio", + "type": "bargauge" + }, + { + "fieldConfig": { + "defaults": { + "decimals": 1, + "mappings": [], + "max": 1, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "orange", + "value": 0.7 + }, + { + "color": "red", + "value": 0.9 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 4, + "x": 4, + "y": 1 + }, + "id": 45, + "options": { + "displayMode": "basic", + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "maxVizHeight": 300, + "minVizHeight": 10, + "minVizWidth": 0, + "namePlacement": "auto", + "orientation": "vertical", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showUnfilled": false, + "sizing": "auto", + "text": {}, + "valueMode": "color" + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "sum (irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",container!=\"\",node=~\"^$Node$\"}[2m])) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\", node=~\"^$Node$\"})", + "format": "time_series", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "CPU Utilization Rate", + "refId": "C" + }, + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",node=~\"^$Node$\"}) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\", node=~\"^$Node$\"})", + "format": "time_series", + "instant": true, + "interval": "", + "legendFormat": "CPU Request Rate", + "refId": "A" + }, + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",node=~\"^$Node$\"}) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\", node=~\"^$Node$\"})", + "format": "time_series", + "instant": true, + "interval": "", + "legendFormat": "CPU Limit Rate", + "refId": "B" + } + ], + "title": "Node CPU Ratio", + "type": "bargauge" + }, + { + "description": "Number of cluster nodes，Nodes POD Number of，NodesPODUpper Limit", + "fieldConfig": { + "defaults": { + "mappings": [], + "max": 100, + "min": 1, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "orange", + "value": 1000 + }, + { + "color": "red", + "value": 2000 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 3, + "x": 8, + "y": 1 + }, + "id": 74, + "options": { + "displayMode": "basic", + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "maxVizHeight": 300, + "minVizHeight": 10, + "minVizWidth": 0, + "namePlacement": "auto", + "orientation": "vertical", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showUnfilled": false, + "sizing": "auto", + "text": {}, + "valueMode": "color" + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "count(kube_node_info{origin_prometheus=~\"$origin_prometheus\"})", + "instant": true, + "interval": "", + "legendFormat": "Number of nodes", + "refId": "A" + }, + { + "expr": "count(kube_pod_info{origin_prometheus=~\"$origin_prometheus\",created_by_kind!~\"|Job\",node=~\"^$Node$\"})", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Pod Number of nodes", + "refId": "B" + }, + { + "exemplar": true, + "expr": "sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"pods\", unit=\"integer\",node=~\"^$Node$\"})", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Upper limit Pod", + "refId": "C" + } + ], + "title": "Nodes with Pod", + "type": "bargauge" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "align": "center", + "cellOptions": { + "type": "color-text" + }, + "filterable": false, + "inspect": false + }, + "mappings": [], + "noValue": "0", + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Space" + }, + "properties": [ + { + "id": "custom.width", + "value": 59 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Pod" + }, + "properties": [ + { + "id": "custom.width", + "value": 21 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "SVC" + }, + "properties": [ + { + "id": "custom.width", + "value": 7 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Microservices" + }, + "properties": [ + { + "id": "custom.width", + "value": 4 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Configuration" + }, + "properties": [ + { + "id": "custom.width", + "value": 16 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Passwords" + }, + "properties": [ + { + "id": "custom.width", + "value": 33 + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 5, + "x": 11, + "y": 1 + }, + "id": 51, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": true + }, + "showHeader": true, + "sortBy": [ + { + "desc": true, + "displayName": "Microservices" + } + ] + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "expr": "count(kube_pod_info{origin_prometheus=~\"$origin_prometheus\",node=~\"^$Node$\"}) by (namespace)", + "format": "table", + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "A" + }, + { + "editorMode": "code", + "expr": "count(kube_service_info{origin_prometheus=~\"$origin_prometheus\"}) by(namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "C" + }, + { + "editorMode": "code", + "expr": "count by (namespace)({__name__=~\"kube_deployment_metadata_generation|kube_daemonset_metadata_generation|kube_statefulset_metadata_generation\",origin_prometheus=~\"$origin_prometheus\"})", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "__auto", + "refId": "D" + }, + { + "editorMode": "code", + "expr": "count(kube_configmap_info{origin_prometheus=~\"$origin_prometheus\"}) by(namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "configmap", + "refId": "B" + }, + { + "editorMode": "code", + "expr": "count(kube_secret_info{origin_prometheus=~\"$origin_prometheus\"}) by(namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "secret", + "refId": "E" + } + ], + "title": "Namespace Resource Statistics", + "transformations": [ + { + "id": "seriesToColumns", + "options": { + "byField": "namespace" + } + }, + { + "id": "organize", + "options": { + "excludeByName": { + "Time": true, + "Time 1": true, + "Time 2": true, + "Time 3": true, + "Time 4": true, + "Time 5": true + }, + "includeByName": {}, + "indexByName": { + "Time 1": 2, + "Time 2": 4, + "Time 3": 6, + "Value #A": 3, + "Value #C": 5, + "Value #D": 1, + "namespace": 0 + }, + "renameByName": { + "Time 1": "", + "Time 2": "", + "Value #A": "Pod", + "Value #B": "Configuration", + "Value #C": "SVC", + "Value #D": "Microservices", + "Value #E": "Passwords", + "namespace": "Spaces" + } + } + } + ], + "type": "table" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "series", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 30, + "gradientMode": "opacity", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineStyle": { + "fill": "solid" + }, + "lineWidth": 2, + "pointSize": 4, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "binbps" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 8, + "x": 16, + "y": 1 + }, + "id": 32, + "options": { + "legend": { + "calcs": ["max"], + "displayMode": "list", + "placement": "bottom", + "showLegend": true, + "width": 200 + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "sum (irate(container_network_receive_bytes_total{origin_prometheus=~\"$origin_prometheus\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m]))*8", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "Receive", + "metric": "network", + "range": true, + "refId": "A", + "step": 10 + }, + { + "editorMode": "code", + "exemplar": true, + "expr": "sum (irate(container_network_transmit_bytes_total{origin_prometheus=~\"$origin_prometheus\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m]))*8", + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "Send", + "metric": "network", + "refId": "B", + "step": 10 + } + ], + "title": "$NameSpace：Network Overview（Associable nodes and namespaces）", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "decimals": 1, + "mappings": [], + "max": 2000000000000, + "min": 1, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "orange", + "value": 100000000000 + }, + { + "color": "red", + "value": 2000000000000 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 4, + "x": 0, + "y": 5 + }, + "id": 71, + "options": { + "displayMode": "basic", + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "maxVizHeight": 300, + "minVizHeight": 10, + "minVizWidth": 0, + "namePlacement": "auto", + "orientation": "vertical", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showUnfilled": false, + "sizing": "auto", + "text": {}, + "valueMode": "color" + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "exemplar": true, + "expr": "sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\", node=~\"^$Node$\"})", + "instant": true, + "interval": "", + "legendFormat": "Total Memory", + "refId": "A" + }, + { + "exemplar": true, + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",container!=\"\",node=~\"^$Node$\"})", + "instant": true, + "interval": "", + "legendFormat": "Usage", + "refId": "C" + }, + { + "exemplar": true, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",node=~\"^$Node$\"})", + "instant": true, + "interval": "", + "legendFormat": "Requests", + "refId": "D" + }, + { + "exemplar": true, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",node=~\"^$Node$\"})", + "instant": true, + "interval": "", + "legendFormat": "Limits", + "refId": "B" + } + ], + "title": "Node Memory Information", + "type": "bargauge" + }, + { + "fieldConfig": { + "defaults": { + "decimals": 1, + "mappings": [], + "max": 500, + "min": 1, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "#EAB839", + "value": 500 + }, + { + "color": "red", + "value": 1000 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 4, + "x": 4, + "y": 5 + }, + "id": 72, + "options": { + "displayMode": "basic", + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "maxVizHeight": 300, + "minVizHeight": 10, + "minVizWidth": 0, + "namePlacement": "auto", + "orientation": "vertical", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showUnfilled": false, + "sizing": "auto", + "text": {}, + "valueMode": "color" + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "exemplar": true, + "expr": "sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\", node=~\"^$Node$\"})", + "instant": true, + "interval": "", + "legendFormat": "Total Cores", + "refId": "A" + }, + { + "exemplar": true, + "expr": "sum (irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",id=\"/\",node=~\"^$Node$\"}[2m]))", + "instant": true, + "interval": "", + "legendFormat": "Usage", + "refId": "C" + }, + { + "exemplar": true, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",node=~\"^$Node$\"})", + "instant": true, + "interval": "", + "legendFormat": "Requests", + "refId": "D" + }, + { + "exemplar": true, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",node=~\"^$Node$\"})", + "instant": true, + "interval": "", + "legendFormat": "Limit", + "refId": "B" + } + ], + "title": "Node CPU Number of cores", + "type": "bargauge" + }, + { + "fieldConfig": { + "defaults": { + "decimals": 1, + "mappings": [], + "max": 8000000000000, + "min": 1, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "#EAB839", + "value": 5000000000000 + }, + { + "color": "red", + "value": 10000000000000 + } + ] + }, + "unit": "bytes" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Utilization" + }, + "properties": [ + { + "id": "unit", + "value": "percentunit" + }, + { + "id": "max" + }, + { + "id": "min", + "value": 0 + }, + { + "id": "thresholds", + "value": { + "mode": "percentage", + "steps": [ + { + "color": "green" + }, + { + "color": "orange", + "value": 80 + }, + { + "color": "red", + "value": 90 + } + ] + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Total" + }, + "properties": [ + { + "id": "decimals", + "value": 0 + } + ] + } + ] + }, + "gridPos": { + "h": 4, + "w": 3, + "x": 8, + "y": 5 + }, + "id": 73, + "options": { + "displayMode": "basic", + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "maxVizHeight": 300, + "minVizHeight": 10, + "minVizWidth": 0, + "namePlacement": "auto", + "orientation": "vertical", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showUnfilled": false, + "sizing": "auto", + "text": {}, + "valueMode": "color" + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum (container_fs_usage_bytes{origin_prometheus=~\"$origin_prometheus\",device=~\"^/dev/.*$\",id=\"/\",node=~\"^$Node$\"}) / sum (container_fs_limit_bytes{origin_prometheus=~\"$origin_prometheus\",device=~\"^/dev/.*$\",id=\"/\",node=~\"^$Node$\"})", + "instant": true, + "interval": "", + "legendFormat": "Utilization Rate", + "refId": "C" + }, + { + "expr": "sum (container_fs_usage_bytes{origin_prometheus=~\"$origin_prometheus\",device=~\"^/dev/.*$\",id=\"/\",node=~\"^$Node$\"})", + "instant": true, + "interval": "", + "legendFormat": "Usage", + "refId": "A" + }, + { + "expr": "sum (container_fs_limit_bytes{origin_prometheus=~\"$origin_prometheus\",device=~\"^/dev/.*$\",id=\"/\",node=~\"^$Node$\"})", + "instant": true, + "interval": "", + "legendFormat": "Total", + "refId": "B" + } + ], + "title": "Node Storage Information", + "type": "bargauge" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/Anomaly.*/" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "red", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 2, + "w": 24, + "x": 0, + "y": 9 + }, + "id": 88, + "maxPerRow": 2, + "options": { + "colorMode": "background", + "graphMode": "none", + "justifyMode": "center", + "orientation": "auto", + "reduceOptions": { + "calcs": ["last"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "text": {}, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "10.4.1", + "repeat": "origin_prometheus", + "repeatDirection": "v", + "targets": [ + { + "editorMode": "code", + "exemplar": false, + "expr": "count({__name__=~\"kube_deployment_metadata_generation|kube_daemonset_metadata_generation|kube_statefulset_metadata_generation\",origin_prometheus=~\"$origin_prometheus\"})", + "hide": false, + "instant": true, + "legendFormat": "Workload", + "range": false, + "refId": "F" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "count(kube_pod_info{origin_prometheus=~\"$origin_prometheus\"})", + "hide": false, + "instant": true, + "legendFormat": "Total Pod", + "range": false, + "refId": "E" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "count by(key,origin_prometheus)(kube_node_spec_taint{origin_prometheus=~\"$origin_prometheus\",key=~\"node.kubernetes.io.*\"})", + "format": "time_series", + "hide": false, + "instant": true, + "legendFormat": "{{key}}", + "range": false, + "refId": "D" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "count by(origin_prometheus)(kube_node_info{origin_prometheus=~\"$origin_prometheus\"})", + "hide": false, + "instant": true, + "legendFormat": "Total Nodes", + "range": false, + "refId": "C" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "count by(origin_prometheus)(kube_node_info{origin_prometheus=~\"$origin_prometheus\"}) - count by(origin_prometheus)(kube_node_spec_taint{origin_prometheus=~\"$origin_prometheus\",key!~\"node.kubernetes.io.*\"})", + "hide": false, + "instant": true, + "legendFormat": "Normal Node", + "range": false, + "refId": "B" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "count by(key,origin_prometheus)(kube_node_spec_taint{origin_prometheus=~\"$origin_prometheus\",key!~\"node.kubernetes.io.*\"})", + "hide": false, + "instant": true, + "legendFormat": "{{key}}", + "range": false, + "refId": "A" + } + ], + "title": "", + "transformations": [ + { + "id": "renameByRegex", + "options": { + "regex": "(node.kubernetes.io/)(.*)", + "renamePattern": "Abnormal:$2" + } + } + ], + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 15, + "gradientMode": "opacity", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Total Memory" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "red", + "mode": "fixed" + } + }, + { + "id": "custom.lineWidth", + "value": 2 + } + ] + } + ] + }, + "gridPos": { + "h": 5, + "w": 8, + "x": 0, + "y": 11 + }, + "id": 79, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\", node=~\"^$Node$\"})", + "instant": false, + "interval": "", + "legendFormat": "Total Memory", + "refId": "A" + }, + { + "exemplar": true, + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",container!=\"\",node=~\"^$Node$\"})", + "instant": false, + "interval": "", + "legendFormat": "Usage", + "refId": "C" + }, + { + "exemplar": true, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",node=~\"^$Node$\"})", + "hide": true, + "instant": false, + "interval": "", + "legendFormat": "Requests", + "refId": "D" + }, + { + "exemplar": true, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",node=~\"^$Node$\"})", + "hide": true, + "instant": false, + "interval": "", + "legendFormat": "Limit", + "refId": "B" + } + ], + "title": "Memory Usage【$Node】", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 15, + "gradientMode": "opacity", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Total Cores" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "red", + "mode": "fixed" + } + }, + { + "id": "custom.lineWidth", + "value": 2 + } + ] + } + ] + }, + "gridPos": { + "h": 5, + "w": 8, + "x": 8, + "y": 11 + }, + "id": 80, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\", node=~\"^$Node$\"})", + "instant": false, + "interval": "", + "legendFormat": "Total Cores", + "refId": "A" + }, + { + "exemplar": true, + "expr": "sum (irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",id=\"/\",node=~\"^$Node$\"}[2m]))", + "instant": false, + "interval": "", + "legendFormat": "Usage", + "refId": "C" + }, + { + "exemplar": true, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",node=~\"^$Node$\"})", + "hide": true, + "instant": false, + "interval": "", + "legendFormat": "Requests", + "refId": "D" + }, + { + "exemplar": true, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",node=~\"^$Node$\"})", + "hide": true, + "instant": true, + "interval": "", + "legendFormat": "Limit", + "refId": "B" + } + ], + "title": "CPU Used Cores【$Node】", + "type": "timeseries" + }, + { + "description": "Number of Cluster Nodes，Nodes POD Number of，Nodes POD Upper Limit", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "series", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 15, + "gradientMode": "opacity", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Upper Limit Pod" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "red", + "mode": "fixed" + } + }, + { + "id": "custom.lineWidth", + "value": 2 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Number of nodes" + }, + "properties": [ + { + "id": "custom.axisPlacement", + "value": "right" + }, + { + "id": "custom.drawStyle", + "value": "points" + }, + { + "id": "color", + "value": { + "fixedColor": "blue", + "mode": "fixed" + } + }, + { + "id": "custom.pointSize", + "value": 3 + } + ] + } + ] + }, + "gridPos": { + "h": 5, + "w": 8, + "x": 16, + "y": 11 + }, + "id": 81, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "exemplar": true, + "expr": "count(kube_node_info{origin_prometheus=~\"$origin_prometheus\"})", + "hide": false, + "instant": false, + "interval": "", + "legendFormat": "Number of nodes", + "refId": "A" + }, + { + "editorMode": "code", + "exemplar": true, + "expr": "count(kube_pod_info{origin_prometheus=~\"$origin_prometheus\",created_by_kind!~\"|Job\",node=~\"^$Node$\"})", + "hide": false, + "instant": false, + "interval": "", + "legendFormat": "Pod Number", + "refId": "B" + }, + { + "exemplar": true, + "expr": "sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"pods\", unit=\"integer\",node=~\"^$Node$\"})", + "hide": false, + "instant": false, + "interval": "", + "legendFormat": "Upper Limit Pod", + "refId": "C" + } + ], + "title": "Pod Number and nodes【$Node】", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/Total number of cores.*/" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#C4162A", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 6, + "w": 8, + "x": 0, + "y": 16 + }, + "id": 75, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "exemplar": true, + "expr": "sum (irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",container!=\"\",node=~\"^$Node$\"}[2m]))by (node) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\", node=~\"^$Node$\"})by (node)*100", + "format": "time_series", + "hide": false, + "instant": false, + "interval": "", + "legendFormat": "{{node}}", + "refId": "I" + } + ], + "title": "$Node：Nodes CPU Breakdown", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 6, + "w": 8, + "x": 8, + "y": 16 + }, + "id": 76, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "exemplar": true, + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",container!=\"\",node=~\"^$Node$\"})by (node) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\", node=~\"^$Node$\"})by (node)*100", + "format": "time_series", + "hide": false, + "instant": false, + "interval": "", + "legendFormat": "{{node}}", + "refId": "I" + } + ], + "title": "$Node：Node Memory Breakdown", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "series", + "axisLabel": "←Inflow/Outflow→", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "binbps" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/Inflow.*/" + }, + "properties": [ + { + "id": "custom.transform", + "value": "negative-Y" + } + ] + } + ] + }, + "gridPos": { + "h": 6, + "w": 8, + "x": 16, + "y": 16 + }, + "id": 78, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "exemplar": true, + "expr": "sum (irate(container_network_receive_bytes_total{origin_prometheus=~\"$origin_prometheus\",node=~\"^$Node$\"}[2m]))by (node) *8", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "Inflow:{{node}}", + "metric": "network", + "refId": "A", + "step": 10 + }, + { + "exemplar": true, + "expr": "sum (irate(container_network_transmit_bytes_total{origin_prometheus=~\"$origin_prometheus\",node=~\"^$Node$\"}[2m]))by (node) *8", + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "Outflows:{{node}}", + "metric": "network", + "refId": "B", + "step": 10 + } + ], + "title": "$Node：Node Network Overview", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "custom": { + "align": "center", + "cellOptions": { + "type": "auto" + }, + "filterable": false, + "inspect": false + }, + "mappings": [], + "max": 1, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "CPU Limitations" + }, + "properties": [ + { + "id": "custom.width", + "value": 76 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Memory Usage" + }, + "properties": [ + { + "id": "custom.width", + "value": 71 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Memory Limits" + }, + "properties": [ + { + "id": "custom.width", + "value": 74 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Disk Usage" + }, + "properties": [ + { + "id": "custom.width", + "value": 74 + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": ".*%" + }, + "properties": [ + { + "id": "custom.cellOptions", + "value": { + "mode": "gradient", + "type": "color-background" + } + }, + { + "id": "color", + "value": { + "mode": "continuous-GrYlRd" + } + }, + { + "id": "custom.width", + "value": 85 + }, + { + "id": "unit", + "value": "percentunit" + }, + { + "id": "decimals", + "value": 0 + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": "(Memory Usage|Total Memory|Memory Request|Memory Limit|Disk Usage|Disk Total)" + }, + "properties": [ + { + "id": "unit", + "value": "bytes" + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Nodes" + }, + "properties": [ + { + "id": "custom.width", + "value": 96 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Memory Requests" + }, + "properties": [ + { + "id": "custom.width", + "value": 76 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "CPU Requests" + }, + "properties": [ + { + "id": "custom.width", + "value": 75 + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": "(CPUTotal|MemoryTotal|DiskTotal|PodLimit)" + }, + "properties": [ + { + "id": "custom.cellOptions", + "value": { + "mode": "gradient", + "type": "color-background" + } + }, + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "blue" + } + ] + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "PodCap" + }, + "properties": [ + { + "id": "custom.width", + "value": 66 + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": "CPU Core Usage$|Memory Usage$|Disk Usage$|Pod Number of" + }, + "properties": [ + { + "id": "custom.cellOptions", + "value": { + "type": "color-text" + } + }, + { + "id": "color", + "value": { + "fixedColor": "orange", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": "/.*Total/" + }, + "properties": [ + { + "id": "custom.cellOptions", + "value": { + "mode": "gradient", + "type": "color-background" + } + }, + { + "id": "custom.cellOptions", + "value": { + "mode": "gradient", + "type": "color-background" + } + }, + { + "id": "color", + "value": { + "fixedColor": "blue", + "mode": "fixed" + } + }, + { + "id": "decimals", + "value": 0 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Pod Number of" + }, + "properties": [ + { + "id": "custom.width", + "value": 58 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "CPUTotal Cores" + }, + "properties": [ + { + "id": "custom.width", + "value": 69 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Total Memory" + }, + "properties": [ + { + "id": "custom.width", + "value": 75 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Total Disk" + }, + "properties": [ + { + "id": "custom.width", + "value": 74 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "CPU Core Usage" + }, + "properties": [ + { + "id": "custom.width", + "value": 74 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Memory Usage%" + }, + "properties": [ + { + "id": "custom.width", + "value": 102 + } + ] + } + ] + }, + "gridPos": { + "h": 11, + "w": 24, + "x": 0, + "y": 22 + }, + "id": 52, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "enablePagination": true, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true, + "sortBy": [ + { + "desc": true, + "displayName": "Memory Usage%" + } + ] + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "exemplar": false, + "expr": "count(kube_pod_info{origin_prometheus=~\"$origin_prometheus\",created_by_kind!~\"|Job\",node=~\"^$Node$\"}) by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "pod Number", + "refId": "A" + }, + { + "exemplar": false, + "expr": "kube_node_status_condition{origin_prometheus=~\"$origin_prometheus\",status=\"true\",node=~\"^$Node$\"} == 1", + "format": "table", + "hide": true, + "instant": true, + "interval": "", + "legendFormat": "Status", + "refId": "B" + }, + { + "exemplar": false, + "expr": "sum (irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",container!=\"\",node=~\"^$Node$\"}[2m])) by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "I" + }, + { + "exemplar": false, + "expr": "kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\", node=~\"^$Node$\"} - 0", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "C" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",node=~\"^$Node$\"}) by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "E" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",node=~\"^$Node$\"}) by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "F" + }, + { + "exemplar": false, + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",container!=\"\",node=~\"^$Node$\"}) by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "J" + }, + { + "exemplar": false, + "expr": "sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\", node=~\"^$Node$\"}) by (node) - 0", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "D" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",node=~\"^$Node$\"}) by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "G" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",node=~\"^$Node$\"}) by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "H" + }, + { + "exemplar": false, + "expr": "sum (container_fs_usage_bytes{origin_prometheus=~\"$origin_prometheus\",device=~\"^/dev/.*$\",id=\"/\",node=~\"^$Node$\"}) by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "K" + }, + { + "exemplar": false, + "expr": "sum (container_fs_limit_bytes{origin_prometheus=~\"$origin_prometheus\",device=~\"^/dev/.*$\",id=\"/\",node=~\"^$Node$\"}) by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "L" + }, + { + "exemplar": false, + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",container!=\"\",node=~\"^$Node$\"})by (node) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\", node=~\"^$Node$\"})by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Memory Usage%", + "refId": "M" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",node=~\"^$Node$\"})by (node) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\", node=~\"^$Node$\"})by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Memory Requests%", + "refId": "N" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",node=~\"^$Node$\"})by (node) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\", node=~\"^$Node$\"})by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Memory Limit%", + "refId": "O" + }, + { + "exemplar": false, + "expr": "sum (irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",container!=\"\",node=~\"^$Node$\"}[2m]))by (node) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\", node=~\"^$Node$\"})by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "CPU Usage%", + "refId": "P" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",node=~\"^$Node$\"})by (node) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\", node=~\"^$Node$\"})by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "CPU Requests%", + "refId": "Q" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",node=~\"^$Node$\"})by (node) / sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\", node=~\"^$Node$\"})by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Memory Limit%", + "refId": "R" + }, + { + "exemplar": false, + "expr": "sum (container_fs_usage_bytes{origin_prometheus=~\"$origin_prometheus\",device=~\"^/dev/.*$\",id=\"/\",node=~\"^$Node$\"})by (node) / sum (container_fs_limit_bytes{origin_prometheus=~\"$origin_prometheus\",device=~\"^/dev/.*$\",id=\"/\",node=~\"^$Node$\"})by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Disk Usage%", + "refId": "S" + }, + { + "exemplar": false, + "expr": "sum(kube_node_status_allocatable{origin_prometheus=~\"$origin_prometheus\",resource=\"pods\", unit=\"integer\",node=~\"^$Node$\"})by (node)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "PodCaps", + "refId": "T" + } + ], + "title": "$Node：Node Information Detail", + "transformations": [ + { + "id": "merge", + "options": {} + }, + { + "id": "organize", + "options": { + "excludeByName": { + "Time": true, + "Time 1": true, + "Time 10": true, + "Time 11": true, + "Time 12": true, + "Time 13": true, + "Time 14": true, + "Time 15": true, + "Time 16": true, + "Time 17": true, + "Time 18": true, + "Time 19": true, + "Time 2": true, + "Time 20": true, + "Time 3": true, + "Time 4": true, + "Time 5": true, + "Time 6": true, + "Time 7": true, + "Time 8": true, + "Time 9": true, + "Value #B": true, + "Value #E": false, + "Value #F": false, + "__name__": true, + "app_kubernetes_io_name": true, + "app_kubernetes_io_name 1": true, + "app_kubernetes_io_name 2": true, + "app_kubernetes_io_name 3": true, + "app_kubernetes_io_version": true, + "app_kubernetes_io_version 1": true, + "app_kubernetes_io_version 2": true, + "app_kubernetes_io_version 3": true, + "condition": true, + "instance": true, + "instance 1": true, + "instance 2": true, + "instance 3": true, + "job": true, + "job 1": true, + "job 2": true, + "job 3": true, + "k8s_namespace": true, + "k8s_namespace 1": true, + "k8s_namespace 2": true, + "k8s_namespace 3": true, + "k8s_sname": true, + "k8s_sname 1": true, + "k8s_sname 2": true, + "k8s_sname 3": true, + "origin_prometheus": true, + "origin_prometheus 1": true, + "origin_prometheus 2": true, + "origin_prometheus 3": true, + "resource": true, + "status": true, + "unit": true + }, + "includeByName": {}, + "indexByName": { + "Time": 22, + "Value #A": 2, + "Value #C": 6, + "Value #D": 8, + "Value #E": 16, + "Value #F": 17, + "Value #G": 18, + "Value #H": 19, + "Value #I": 7, + "Value #J": 9, + "Value #K": 11, + "Value #L": 10, + "Value #M": 4, + "Value #N": 13, + "Value #O": 15, + "Value #P": 3, + "Value #Q": 12, + "Value #R": 14, + "Value #S": 5, + "Value #T": 1, + "instance": 23, + "job": 24, + "node": 0, + "origin_prometheus": 25, + "resource": 20, + "unit": 21 + }, + "renameByName": { + "Value #A": "Pod Number of", + "Value #C": "CPU Total Cores", + "Value #D": "Total Memory", + "Value #E": "CPU Requests", + "Value #F": "CPU Limit", + "Value #G": "Memory Requests", + "Value #H": "Memory Limit", + "Value #I": "CPU Core Usage", + "Value #J": "Memory Usage", + "Value #K": "Disk Usage", + "Value #L": "Disk Total", + "Value #M": "Memory Usage%", + "Value #N": "Memory Requests%", + "Value #O": "Memory Limit%", + "Value #P": "CPU Usage%", + "Value #Q": "CPU Requests%", + "Value #R": "CPU Limit%", + "Value #S": "Disk Usage%", + "Value #T": "Pod Limit", + "condition": "Status", + "node": "Node" + } + } + }, + { + "id": "filterFieldsByName", + "options": {} + } + ], + "type": "table" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "align": "center", + "cellOptions": { + "type": "color-background" + }, + "inspect": false + }, + "decimals": 0, + "links": [], + "mappings": [ + { + "options": { + "0": { + "color": "red", + "index": 0 + } + }, + "type": "value" + }, + { + "options": { + "match": "null", + "result": { + "color": "red", + "index": 1 + } + }, + "type": "special" + } + ], + "noValue": "0", + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + } + ] + }, + "unit": "bytes" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Utilization" + }, + "properties": [ + { + "id": "unit", + "value": "percent" + }, + { + "id": "custom.width", + "value": 54 + }, + { + "id": "color", + "value": { + "fixedColor": "purple", + "mode": "fixed" + } + }, + { + "id": "decimals" + }, + { + "id": "mappings", + "value": [ + { + "options": { + "from": 75, + "result": { + "color": "semi-dark-red", + "index": 0 + }, + "to": 110 + }, + "type": "range" + } + ] + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Mounts Pod Number of" + }, + "properties": [ + { + "id": "unit", + "value": "none" + }, + { + "id": "custom.width", + "value": 59 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Namespace" + }, + "properties": [ + { + "id": "custom.width", + "value": 58 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "PVC" + }, + "properties": [ + { + "id": "custom.width", + "value": 94 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Usage" + }, + "properties": [ + { + "id": "custom.width", + "value": 57 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Total" + }, + "properties": [ + { + "id": "custom.width", + "value": 54 + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 6, + "x": 0, + "y": 33 + }, + "id": 92, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true, + "sortBy": [ + { + "desc": true, + "displayName": "Usage Rate" + } + ] + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": false, + "expr": "max by (namespace,persistentvolumeclaim) (kubelet_volume_stats_used_bytes{origin_prometheus=~\"$origin_prometheus\"})", + "format": "table", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{namespace}}:{{ persistentvolumeclaim }}", + "metric": "container_memory_usage:sort_desc", + "range": false, + "refId": "A", + "step": 10 + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "min by (namespace,persistentvolumeclaim) (kubelet_volume_stats_available_bytes{origin_prometheus=~\"$origin_prometheus\"}) + max by (namespace,persistentvolumeclaim) (kubelet_volume_stats_used_bytes{origin_prometheus=~\"$origin_prometheus\"})", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "metric": "container_memory_usage:sort_desc", + "range": false, + "refId": "B", + "step": 10 + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "max by (namespace,persistentvolumeclaim) (kubelet_volume_stats_used_bytes{origin_prometheus=~\"$origin_prometheus\"}) /(min by (namespace,persistentvolumeclaim) (kubelet_volume_stats_available_bytes{origin_prometheus=~\"$origin_prometheus\"}) + max by (namespace,persistentvolumeclaim) (kubelet_volume_stats_used_bytes{origin_prometheus=~\"$origin_prometheus\"}))*100", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{namespace}}:{{ persistentvolumeclaim }}", + "metric": "container_memory_usage:sort_desc", + "range": false, + "refId": "C", + "step": 10 + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "count by (namespace,persistentvolumeclaim)(kube_pod_spec_volumes_persistentvolumeclaims_info{origin_prometheus=~\"$origin_prometheus\"})", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "metric": "container_memory_usage:sort_desc", + "range": false, + "refId": "D", + "step": 10 + } + ], + "title": "PVC Storage Usage", + "transformations": [ + { + "id": "merge", + "options": {} + }, + { + "id": "organize", + "options": { + "excludeByName": { + "Time": true + }, + "includeByName": {}, + "indexByName": {}, + "renameByName": { + "Value #A": "Usage", + "Value #B": "Total", + "Value #C": "Usage Rate", + "Value #D": "Mount Pod Number", + "namespace": "Namespaces", + "persistentvolumeclaim": "PVC" + } + } + } + ], + "type": "table" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 9, + "x": 6, + "y": 33 + }, + "id": 86, + "options": { + "legend": { + "calcs": ["last", "max"], + "displayMode": "table", + "placement": "right", + "showLegend": true, + "sortBy": "Max", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",container !=\"\",container!=\"POD\"}[2m])) by (namespace)>0.5", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{ namespace }}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + } + ], + "title": "Namespaces CPU Usage kernel(>0.5)", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 9, + "x": 15, + "y": 33 + }, + "id": 85, + "options": { + "legend": { + "calcs": ["last", "max"], + "displayMode": "table", + "placement": "right", + "showLegend": true, + "sortBy": "Max", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",container !=\"\",container!=\"POD\"}) by (namespace) > 1*1024*1024*1024", + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{namespace} {{ pod }}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "A", + "step": 10 + } + ], + "title": "Namespaces WSS Memory Usage (>1G)", + "type": "timeseries" + }, + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 41 + }, + "id": 49, + "panels": [], + "title": "Pod Resource Overview：SelectedPod:【$Pod】", + "type": "row" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "custom": { + "align": "center", + "cellOptions": { + "type": "auto" + }, + "filterable": false, + "inspect": false + }, + "displayName": "", + "mappings": [], + "max": 100, + "min": 0, + "thresholds": { + "mode": "percentage", + "steps": [ + { + "color": "green" + }, + { + "color": "#EAB839", + "value": 80 + }, + { + "color": "red", + "value": 90 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Namespace" + }, + "properties": [ + { + "id": "custom.width", + "value": 96 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Pod Name" + }, + "properties": [ + { + "id": "custom.width", + "value": 207 + }, + { + "id": "custom.align", + "value": "right" + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Cores Used" + }, + "properties": [ + { + "id": "custom.width", + "value": 71 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Reboot" + }, + "properties": [ + { + "id": "custom.width", + "value": 38 + }, + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "#EAB839", + "value": 1 + }, + { + "color": "red", + "value": 3 + } + ] + } + }, + { + "id": "custom.cellOptions", + "value": { + "mode": "gradient", + "type": "color-background" + } + }, + { + "id": "decimals" + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": ".*%" + }, + "properties": [ + { + "id": "unit", + "value": "percentunit" + }, + { + "id": "custom.cellOptions", + "value": { + "type": "color-background" + } + }, + { + "id": "color", + "value": { + "mode": "continuous-GrYlRd" + } + }, + { + "id": "decimals", + "value": 1 + }, + { + "id": "custom.width", + "value": 55 + }, + { + "id": "min", + "value": 0 + }, + { + "id": "max", + "value": 1 + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": ".*Restrictions" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "blue", + "mode": "fixed" + } + }, + { + "id": "custom.cellOptions", + "value": { + "mode": "gradient", + "type": "color-background" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Nodes" + }, + "properties": [ + { + "id": "custom.width", + "value": 100 + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": "Cores in use$|WSS$|RSS$|Survival|Inflow|Outflow" + }, + "properties": [ + { + "id": "custom.cellOptions", + "value": { + "type": "color-text" + } + }, + { + "id": "color", + "value": { + "fixedColor": "orange", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Container Name" + }, + "properties": [ + { + "id": "custom.width", + "value": 57 + }, + { + "id": "custom.align", + "value": "left" + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Survival" + }, + "properties": [ + { + "id": "unit", + "value": "s" + }, + { + "id": "custom.width", + "value": 80 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Used cores" + }, + "properties": [ + { + "id": "custom.width", + "value": 62 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "CPU Limits" + }, + "properties": [ + { + "id": "custom.width", + "value": 58 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Memory Limit" + }, + "properties": [ + { + "id": "custom.width", + "value": 68 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Memory Requirements" + }, + "properties": [ + { + "id": "custom.width", + "value": 88 + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": "WSS$|RSS$|Memory Requirements$|Memory Limit$|Disk.*$" + }, + "properties": [ + { + "id": "unit", + "value": "bytes" + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "WSS" + }, + "properties": [ + { + "id": "custom.width", + "value": 81 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "RSS" + }, + "properties": [ + { + "id": "custom.width", + "value": 74 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "CPU Requirements" + }, + "properties": [ + { + "id": "custom.width", + "value": 72 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Disk Limitations" + }, + "properties": [ + { + "id": "custom.width", + "value": 83 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Disk Usage" + }, + "properties": [ + { + "id": "custom.width", + "value": 72 + }, + { + "id": "custom.cellOptions", + "value": { + "type": "color-background" + } + }, + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "orange", + "value": 10737418240 + }, + { + "color": "red", + "value": 16106127360 + } + ] + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "WSS%" + }, + "properties": [ + { + "id": "custom.width", + "value": 77 + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": "/Inflow|Streaming Out/" + }, + "properties": [ + { + "id": "unit", + "value": "binbps" + }, + { + "id": "custom.width", + "value": 80 + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 24, + "x": 0, + "y": 42 + }, + "id": 47, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true, + "sortBy": [ + { + "desc": true, + "displayName": "Container Name" + } + ] + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": false, + "expr": "sum(irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m])) by (container, pod,node,namespace) / (sum(container_spec_cpu_quota{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}/100000) by (container, pod,node,namespace)) ", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "A" + }, + { + "exemplar": false, + "expr": "sum(irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m])) by (container, pod,node,namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "CPUKernel Usage", + "refId": "Q" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",pod=~\"$Pod\",container =~\"$Container\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container,pod,node,namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "B" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",pod=~\"$Pod\",container =~\"$Container\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container,pod,node,namespace)", + "format": "table", + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "C" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod,node,namespace)/ sum(container_spec_memory_limit_bytes{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod,node,namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "wss%", + "refId": "I" + }, + { + "exemplar": false, + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod,node,namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "wss", + "refId": "D" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "sum (container_memory_rss{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod,node,namespace)/ sum(container_spec_memory_limit_bytes{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod,node,namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "rss%", + "refId": "L" + }, + { + "exemplar": false, + "expr": "sum (container_memory_rss{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod,node,namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "rss", + "refId": "K" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",pod=~\"$Pod\",container =~\"$Container\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container,pod,node,namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "E" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",pod=~\"$Pod\",container =~\"$Container\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container,pod,node,namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "F" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "sum(container_fs_usage_bytes{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container,pod,node,namespace)", + "format": "table", + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "J" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "kube_pod_container_status_restarts_total{\n origin_prometheus=~\"$origin_prometheus\",\n pod=~\"$Pod\",\n container=~\"$Container\",\n container!=\"\",\n container!=\"POD\",\n namespace=~\"$NameSpace\"\n}\n* on (pod) group_left(node) (\n kube_pod_info{\n origin_prometheus=~\"$origin_prometheus\",\n pod=~\"$Pod\",\n node=~\"^$Node$\",\n namespace=~\"$NameSpace\"\n }\n unless on (pod) (\n kube_pod_info{\n origin_prometheus=~\"$origin_prometheus\",\n pod=~\"$Pod\",\n node=~\"^$Node$\",\n namespace=~\"$NameSpace\"\n } offset 1m\n )\n)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "H" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "time() - \n(\n max by(pod, namespace) (kube_pod_created{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",namespace=~\"$NameSpace\"})\n * on(pod, namespace) group_right(container)\n kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",namespace=~\"$NameSpace\",container=~\"$Container\"}\n)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "__auto", + "refId": "R" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "sum(container_fs_limit_bytes{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container,pod,node,namespace)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "S" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "sum(sum(irate(container_network_receive_bytes_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m])) by (pod)* on(pod) group_right kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",namespace=~\"$NameSpace\",container =~\"$Container\"}) by(pod) *8", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "__auto", + "refId": "T" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "sum(sum(irate(container_network_transmit_bytes_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m])) by (pod)* on(pod) group_right kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",namespace=~\"$NameSpace\",container =~\"$Container\"}) by(pod) *8", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "__auto", + "refId": "U" + } + ], + "title": "$Node：Pod Resource Detail(Associable Nodes)", + "transformations": [ + { + "id": "merge", + "options": {} + }, + { + "id": "organize", + "options": { + "excludeByName": { + "Time": true, + "Time 1": true, + "Time 10": true, + "Time 11": true, + "Time 12": true, + "Time 13": true, + "Time 2": true, + "Time 3": true, + "Time 4": true, + "Time 5": true, + "Time 6": true, + "Time 7": true, + "Time 8": true, + "Time 9": true, + "Value #G": false, + "__name__": true, + "app_kubernetes_io_name": true, + "app_kubernetes_io_name 1": true, + "app_kubernetes_io_name 2": true, + "app_kubernetes_io_version": true, + "app_kubernetes_io_version 1": true, + "app_kubernetes_io_version 2": true, + "container 1": true, + "container 10": true, + "container 11": true, + "container 12": true, + "container 2": true, + "container 3": true, + "container 4": true, + "container 5": true, + "container 6": true, + "container 7": true, + "container 8": true, + "container 9": true, + "created_by_kind": true, + "created_by_name": true, + "host_ip": true, + "instance": true, + "instance 1": true, + "instance 2": true, + "job": true, + "job 1": true, + "job 2": true, + "k8s_namespace": true, + "k8s_namespace 1": true, + "k8s_namespace 2": true, + "k8s_sname": true, + "k8s_sname 1": true, + "k8s_sname 2": true, + "namespace": false, + "namespace 1": true, + "namespace 10": true, + "namespace 11": true, + "namespace 12": false, + "namespace 2": true, + "namespace 3": true, + "namespace 4": true, + "namespace 5": true, + "namespace 6": true, + "namespace 7": true, + "namespace 8": true, + "namespace 9": true, + "node 1": true, + "node 10": true, + "node 11": false, + "node 12": true, + "node 2": true, + "node 3": true, + "node 4": true, + "node 5": true, + "node 6": true, + "node 7": true, + "node 8": true, + "node 9": true, + "origin_prometheus": true, + "origin_prometheus 1": true, + "origin_prometheus 2": true, + "phase": true, + "pod_ip": true, + "priority_class": true, + "uid": true + }, + "includeByName": {}, + "indexByName": { + "Time": 21, + "Value #A": 4, + "Value #B": 16, + "Value #C": 7, + "Value #D": 10, + "Value #E": 17, + "Value #F": 9, + "Value #G": 23, + "Value #H": 14, + "Value #I": 5, + "Value #J": 13, + "Value #K": 11, + "Value #L": 6, + "Value #M": 24, + "Value #N": 25, + "Value #O": 26, + "Value #P": 27, + "Value #Q": 8, + "Value #R": 15, + "Value #S": 12, + "container": 2, + "instance": 18, + "ip": 28, + "job": 19, + "namespace": 1, + "node": 0, + "origin_prometheus": 20, + "pod": 3, + "uid": 22 + }, + "renameByName": { + "Value #A": "CPU%", + "Value #B": "CPU Requirements", + "Value #C": "CPU Limitations", + "Value #D": "WSS", + "Value #E": "Memory Requirements", + "Value #F": "Memory Limit", + "Value #H": "Reboot", + "Value #I": "WSS%", + "Value #J": "Disk Usage", + "Value #K": "RSS", + "Value #L": "RSS%", + "Value #M": "Heap Memory", + "Value #N": "maxHeap", + "Value #O": "Non-Heap", + "Value #P": "maxNon-Heap", + "Value #Q": "Using Cores", + "Value #R": "Survival", + "Value #S": "Disk Limit", + "Value #T": "Inflow", + "Value #U": "Streaming Out", + "container": "Container Name", + "instance": "", + "ip": "POD IP", + "namespace": "Namespace", + "namespace 1": "", + "namespace 12": "Namespace", + "node": "Nodes", + "node 1": "", + "node 11": "Node", + "pod": "Pod Name", + "priority_class": "" + } + } + }, + { + "id": "filterFieldsByName", + "options": { + "include": { + "names": [ + "Node", + "Namespace", + "Container Name", + "Pod Name", + "CPU%", + "WSS%", + "RSS%", + "CPU Restrictions", + "Cores Used", + "Memory Limit", + "WSS", + "RSS", + "Disk Limit", + "Disk Usage", + "Reboot", + "CPUDemand", + "Memory Requirements", + "Inflow", + "Outflow", + "Survival" + ] + } + } + } + ], + "type": "table" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 8, + "x": 0, + "y": 50 + }, + "id": 58, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "max(irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m])) by (container, pod) / (max(container_spec_cpu_quota{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}/100000) by (container, pod)) * 100", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{ pod }}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + } + ], + "title": "Pod Containers CPU Utilization (Maximum100%Associable Nodes)", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 8, + "x": 8, + "y": 50 + }, + "id": 27, + "options": { + "legend": { + "calcs": ["max", "last", "mean"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": false, + "expr": "max (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod)/ max(container_spec_memory_limit_bytes{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod) * 100", + "interval": "", + "intervalFactor": 1, + "legendFormat": "WSS：{{ pod }}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "A", + "step": 10 + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "max (container_memory_rss{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod)/ max(container_spec_memory_limit_bytes{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod) * 100", + "interval": "", + "intervalFactor": 1, + "legendFormat": "RSS：{{ pod }}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "B", + "step": 10 + }, + { + "editorMode": "code", + "exemplar": true, + "expr": "(cass_jvm_heap{service=~\"$Container\"} * on (pod_ip) group_right(service) kube_pod_info{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",namespace=~\"$NameSpace\"}) / (cass_jvm_heap_max{service=~\"$Container\"} * on (pod_ip) group_right(service) kube_pod_info{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",namespace=~\"$NameSpace\"}) * 100", + "hide": true, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "Heap：{{ pod }}", + "metric": "container_memory_usage:sort_desc", + "refId": "C", + "step": 10 + } + ], + "title": "Pod Container Memory Usage (Associatable Nodes)", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "binbps" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 8, + "x": 16, + "y": 50 + }, + "id": 77, + "options": { + "legend": { + "calcs": ["max", "last", "mean"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": false, + "expr": "max(max(irate(container_network_receive_bytes_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m])) by (pod)* on(pod) group_right kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",namespace=~\"$NameSpace\",container =~\"$Container\"}) by(pod) *8", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "Inflow:{{ pod}}", + "metric": "network", + "range": true, + "refId": "A", + "step": 10 + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "max(max(irate(container_network_transmit_bytes_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m])) by (pod)* on(pod) group_right kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",namespace=~\"$NameSpace\",container =~\"$Container\"}) by(pod) *8", + "interval": "", + "intervalFactor": 1, + "legendFormat": "Outflow:{{ pod}}", + "metric": "network", + "range": true, + "refId": "B", + "step": 10 + }, + { + "editorMode": "code", + "expr": "sum(sum(irate(container_network_receive_bytes_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",image!=\"\",name=~\"^k8s_.*\",node=~\"^$Node$\",namespace=~\"$NameSpace\",pod=~\".*$Container.*\"}[2m])) by (pod)* on(pod) group_right kube_pod_container_info) by(pod) *8", + "hide": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "metric": "network", + "range": true, + "refId": "C", + "step": 10 + }, + { + "editorMode": "code", + "expr": "sum(sum(irate(container_network_transmit_bytes_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",image!=\"\",name=~\"^k8s_.*\",node=~\"^$Node$\",namespace=~\"$NameSpace\",pod=~\".*$Container.*\"}[2m])) by (pod)* on(pod) group_right kube_pod_container_info) by(pod) *8", + "hide": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "__auto", + "metric": "network", + "range": true, + "refId": "D", + "step": 10 + } + ], + "title": "Pod Network bandwidth per second (Associable Nodes)", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/Limit.*/" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "red", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 9, + "w": 8, + "x": 0, + "y": 59 + }, + "id": 82, + "options": { + "legend": { + "calcs": ["max", "last", "mean"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Max", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "max(irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m])) by (container, pod,node,namespace)", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "CPU Usage：{{ pod }}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + }, + { + "editorMode": "code", + "exemplar": true, + "expr": "max(max(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",pod=~\"$Pod\",container =~\"$Container\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container,pod,node,namespace)) by(container)", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "Pod CPU Limitations：{{ container}}", + "metric": "container_cpu", + "refId": "B", + "step": 10 + } + ], + "title": "Pod Containers CPU Core Usage", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/.*Limitations/" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "red", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 9, + "w": 8, + "x": 8, + "y": 59 + }, + "id": 84, + "options": { + "legend": { + "calcs": ["max", "last", "mean"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Max", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": false, + "expr": "max (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod,node,namespace)", + "interval": "", + "intervalFactor": 1, + "legendFormat": "WSS：{{ pod }}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "A", + "step": 10 + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "max(max(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",pod=~\"$Pod\",container =~\"$Container\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container,pod,node,namespace)) by(container)", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "PodMemory Limits：{{ container}}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "B", + "step": 10 + } + ], + "title": "Pod Containers WSS Memory Usage (Associable Nodes)", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 8, + "x": 16, + "y": 59 + }, + "id": 83, + "options": { + "legend": { + "calcs": ["lastNotNull", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "max (container_memory_rss{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",container =~\"$Container\",container !=\"\",container!=\"POD\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}) by (container, pod,node,namespace)", + "interval": "", + "intervalFactor": 1, + "legendFormat": "RSS：{{ pod }}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "A", + "step": 10 + } + ], + "title": "Pod Containers RSS Memory Usage (Associable Nodes)", + "type": "timeseries" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 68 + }, + "id": 61, + "panels": [ + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "custom": { + "align": "center", + "cellOptions": { + "type": "auto" + }, + "filterable": false, + "inspect": false + }, + "displayName": "", + "mappings": [], + "thresholds": { + "mode": "percentage", + "steps": [ + { + "color": "green" + } + ] + }, + "unit": "none" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": ".*%" + }, + "properties": [ + { + "id": "unit", + "value": "percentunit" + }, + { + "id": "custom.cellOptions", + "value": { + "mode": "gradient", + "type": "gauge", + "valueDisplayMode": "color" + } + }, + { + "id": "max", + "value": 1 + }, + { + "id": "min", + "value": 0 + }, + { + "id": "color", + "value": { + "mode": "continuous-GrYlRd" + } + }, + { + "id": "decimals", + "value": 0 + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": ".*Memory Usage$|.*Memory Limits$|.*Memory Requirements$|.*Disk Usage$|.*Disk Limits$" + }, + "properties": [ + { + "id": "unit", + "value": "bytes" + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Namespace" + }, + "properties": [ + { + "id": "custom.width", + "value": 92 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Container Name" + }, + "properties": [ + { + "id": "custom.width", + "value": 187 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Total CPU Core Usage" + }, + "properties": [ + { + "id": "custom.width", + "value": 100 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Pod" + }, + "properties": [ + { + "id": "custom.width", + "value": 44 + }, + { + "id": "custom.cellOptions", + "value": { + "type": "color-background" + } + }, + { + "id": "color", + "value": { + "fixedColor": "blue", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Average CPU Usage%" + }, + "properties": [ + { + "id": "custom.width", + "value": 116 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Average RSS Memory Usage%" + }, + "properties": [ + { + "id": "custom.width", + "value": 141 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Average WSS Memory Usage%" + }, + "properties": [ + { + "id": "custom.width", + "value": 165 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Total CPU Limit" + }, + "properties": [ + { + "id": "custom.width", + "value": 86 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Total Memory Limit" + }, + "properties": [ + { + "id": "custom.width", + "value": 86 + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": "/.*Limit$/" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "blue", + "mode": "fixed" + } + }, + { + "id": "custom.cellOptions", + "value": { + "type": "color-background" + } + } + ] + }, + { + "matcher": { + "id": "byRegexp", + "options": "/.*Memory Usage$|.*Core Usage$/" + }, + "properties": [ + { + "id": "custom.cellOptions", + "value": { + "type": "color-text" + } + }, + { + "id": "color", + "value": { + "fixedColor": "orange", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Total RSS Memory Usage" + }, + "properties": [ + { + "id": "custom.width", + "value": 107 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Total WSS Memory Usage" + }, + "properties": [ + { + "id": "custom.width", + "value": 113 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Average Disk Usage" + }, + "properties": [ + { + "id": "custom.width", + "value": 96 + }, + { + "id": "custom.cellOptions", + "value": { + "type": "color-background" + } + }, + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "orange", + "value": 10737418240 + }, + { + "color": "red", + "value": 16106127360 + } + ] + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Average Disk Limit" + }, + "properties": [ + { + "id": "custom.width", + "value": 96 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Total CPU Demand" + }, + "properties": [ + { + "id": "custom.width", + "value": 80 + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Total Memory Demand" + }, + "properties": [ + { + "id": "custom.width", + "value": 80 + } + ] + } + ] + }, + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 3 + }, + "id": 87, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "enablePagination": true, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true, + "sortBy": [ + { + "desc": true, + "displayName": "Average WSS Memory Usage%" + } + ] + }, + "pluginVersion": "10.4.1", + "targets": [ + { + "editorMode": "code", + "exemplar": false, + "expr": "sum(irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}[2m])) by (container) / (sum(container_spec_cpu_quota{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}/100000) by (container))", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "A" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "sum(irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}[2m])) by (container)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Total Cores Used", + "refId": "L" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "B" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "format": "table", + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "C" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)/ sum(container_spec_memory_limit_bytes{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "I" + }, + { + "exemplar": false, + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "D" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "sum (container_memory_rss{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)/ sum(container_spec_memory_limit_bytes{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Average Memory%(RSS)", + "refId": "H" + }, + { + "exemplar": false, + "expr": "sum (container_memory_rss{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "Total Memory Usage (RSS) ", + "refId": "K" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "sum by(container) (kube_pod_container_resource_requests{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"})", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "E" + }, + { + "exemplar": false, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"memory\", unit=\"byte\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "F" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "avg(container_fs_usage_bytes{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "format": "table", + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "J" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "count(kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by(container,namespace)", + "format": "table", + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "G" + }, + { + "editorMode": "code", + "exemplar": false, + "expr": "avg(container_fs_limit_bytes{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "legendFormat": "", + "refId": "M" + } + ], + "title": "Microservices (Container Name) Resource Statistics", + "transformations": [ + { + "id": "merge", + "options": {} + }, + { + "id": "organize", + "options": { + "excludeByName": { + "Time": true, + "Time 1": true, + "Time 10": true, + "Time 11": true, + "Time 12": true, + "Time 2": true, + "Time 3": true, + "Time 4": true, + "Time 5": true, + "Time 6": true, + "Time 7": true, + "Time 8": true, + "Time 9": true + }, + "includeByName": {}, + "indexByName": { + "Time": 15, + "Value #A": 3, + "Value #B": 13, + "Value #C": 6, + "Value #D": 9, + "Value #E": 14, + "Value #F": 8, + "Value #G": 2, + "Value #H": 5, + "Value #I": 4, + "Value #J": 12, + "Value #K": 10, + "Value #L": 7, + "Value #M": 11, + "container": 1, + "namespace": 0 + }, + "renameByName": { + "Time 1": "", + "Value #A": "Average CPU Usage%", + "Value #B": "Total CPU Demand", + "Value #C": "Total CPU Restrictions", + "Value #D": "Total WSS Memory Usage", + "Value #E": "Total Memory Requirements", + "Value #F": "Total Memory Limit", + "Value #G": "Pod", + "Value #H": "Average RSS Memory Usage%", + "Value #I": "Average WSS Memory Usage%", + "Value #J": "Average Disk Usage", + "Value #K": "Total RSS Memory Use", + "Value #L": "Total CPU Core Usage", + "Value #M": "Average Disk Limit", + "container": "Container Name", + "namespace": "Namespace" + } + } + }, + { + "id": "filterFieldsByName", + "options": {} + } + ], + "type": "table" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 8, + "x": 0, + "y": 12 + }, + "id": 24, + "options": { + "legend": { + "calcs": ["max", "last", "mean"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "10.4.1", + "targets": [ + { + "exemplar": true, + "expr": "sum(irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}[2m])) by (container) / (sum(container_spec_cpu_quota{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}/100000) by (container)) * 100", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{ container}}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + } + ], + "title": "Microservices (Container Name) Average CPU Usage (Maximum100%)", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 8, + "x": 8, + "y": 12 + }, + "id": 89, + "options": { + "legend": { + "calcs": ["max", "last", "mean"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "10.4.1", + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)/ sum(container_spec_memory_limit_bytes{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container) * 100", + "interval": "", + "intervalFactor": 1, + "legendFormat": "WSS：{{ container }}", + "metric": "container_memory_usage:sort_desc", + "refId": "A", + "step": 10 + }, + { + "expr": "sum (container_memory_rss{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)/ sum(container_spec_memory_limit_bytes{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container) * 100", + "interval": "", + "intervalFactor": 1, + "legendFormat": "RSS：{{ container }}", + "metric": "container_memory_usage:sort_desc", + "refId": "B", + "step": 10 + } + ], + "title": "Microservice (Container Name) Average Memory Utilization", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "binbps" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 8, + "x": 16, + "y": 12 + }, + "id": 16, + "options": { + "legend": { + "calcs": ["mean", "last", "max"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "10.4.1", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(sum(irate(container_network_receive_bytes_total{origin_prometheus=~\"$origin_prometheus\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m])) by (pod)* on(pod) group_right kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",namespace=~\"$NameSpace\",container =~\"$Container\"}) by(container) *8", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "Inflow:{{ container }}", + "metric": "network", + "range": true, + "refId": "A", + "step": 10 + }, + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(sum(irate(container_network_transmit_bytes_total{origin_prometheus=~\"$origin_prometheus\",node=~\"^$Node$\",namespace=~\"$NameSpace\"}[2m])) by (pod)* on(pod) group_right kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",namespace=~\"$NameSpace\",container =~\"$Container\"}) by(container) *8", + "interval": "", + "intervalFactor": 1, + "legendFormat": "Outflow:{{ container }}", + "metric": "network", + "range": true, + "refId": "B", + "step": 10 + }, + { + "expr": "sum (rate (container_network_receive_bytes_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",image!=\"\",name=~\"^k8s_.*\",node=~\"^$Node$\",namespace=~\"$NameSpace\",pod=~\".*$Container.*\"}[2m])) by (pod)", + "hide": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "-> {{ pod }}", + "metric": "network", + "refId": "C", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{origin_prometheus=~\"$origin_prometheus\",pod=~\"$Pod\",image!=\"\",name=~\"^k8s_.*\",node=~\"^$Node$\",namespace=~\"$NameSpace\",pod=~\".*$Container.*\"}[2m])) by (pod)", + "hide": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "<- {{ pod }}", + "metric": "network", + "refId": "D", + "step": 10 + } + ], + "title": "Microservice (Container Name) Network bandwidth per second (Associable Nodes)", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/CPU Limit.*/" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "red", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 8, + "x": 0, + "y": 19 + }, + "id": 91, + "options": { + "legend": { + "calcs": ["max", "last", "mean"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Max", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "10.4.1", + "targets": [ + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(kube_pod_container_resource_limits{origin_prometheus=~\"$origin_prometheus\",resource=\"cpu\", unit=\"core\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "CPU Limitations：{{ container}}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + }, + { + "editorMode": "code", + "exemplar": true, + "expr": "sum(irate(container_cpu_usage_seconds_total{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}[2m])) by (container)", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "CPU Core Usage：{{ container}}", + "metric": "container_cpu", + "refId": "B", + "step": 10 + } + ], + "title": "Microservices (Container Name) Overall CPU Cores Used", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [ + { + "matcher": { + "id": "byRegexp", + "options": "/Memory Limit.*/" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "red", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 8, + "x": 8, + "y": 19 + }, + "id": 90, + "options": { + "legend": { + "calcs": ["max", "last", "mean"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true, + "sortBy": "Max", + "sortDesc": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "10.4.1", + "targets": [ + { + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "interval": "", + "intervalFactor": 1, + "legendFormat": "WSS：{{ container }}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "A", + "step": 10 + }, + { + "editorMode": "code", + "expr": "sum(container_spec_memory_limit_bytes{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "interval": "", + "intervalFactor": 1, + "legendFormat": "Memory Limit：{{ container }}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "B", + "step": 10 + }, + { + "editorMode": "code", + "expr": "sum (container_memory_rss{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by (container)", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "RSS：{{ container }}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "C", + "step": 10 + } + ], + "title": "Microservices (Container Name) Overall Memory Usage", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 8, + "x": 16, + "y": 19 + }, + "id": 59, + "options": { + "legend": { + "calcs": ["max", "last", "mean"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "mode": "multi", + "sort": "desc" + } + }, + "pluginVersion": "10.4.1", + "targets": [ + { + "editorMode": "code", + "expr": "count(kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",container =~\"$Container\",container !=\"\",container!=\"POD\",namespace=~\"$NameSpace\"}) by(container,namespace)", + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{namespace}}：{{ container }}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "A", + "step": 10 + } + ], + "title": "Microservice (Container Name) Pod Number", + "type": "timeseries" + } + ], + "title": "Microservices (Container Name) Resource Overview：Selected Microservices:【$Container】", + "type": "row" + } + ], + "preload": false, + "refresh": "", + "schemaVersion": 41, + "tags": ["Prometheus", "Kubernetes"], + "templating": { + "list": [ + { + "current": { + "text": "", + "value": "" + }, + "definition": "label_values(kube_node_info,origin_prometheus)", + "includeAll": false, + "label": "K8S", + "name": "origin_prometheus", + "options": [], + "query": { + "query": "label_values(kube_node_info,origin_prometheus)", + "refId": "StandardVariableQuery" + }, + "refresh": 2, + "regex": "", + "sort": 5, + "type": "query" + }, + { + "allValue": ".*", + "current": { + "text": "All", + "value": "$__all" + }, + "definition": "label_values(kube_node_info{origin_prometheus=~\"$origin_prometheus\"},node)", + "includeAll": true, + "label": "Nodes", + "name": "Node", + "options": [], + "query": { + "query": "label_values(kube_node_info{origin_prometheus=~\"$origin_prometheus\"},node)", + "refId": "StandardVariableQuery" + }, + "refresh": 2, + "regex": "", + "sort": 5, + "type": "query" + }, + { + "allValue": ".*", + "current": { + "text": "All", + "value": "$__all" + }, + "definition": "label_values(kube_namespace_created{origin_prometheus=~\"$origin_prometheus\"},namespace)", + "includeAll": true, + "label": "Namespace", + "name": "NameSpace", + "options": [], + "query": { + "query": "label_values(kube_namespace_created{origin_prometheus=~\"$origin_prometheus\"},namespace)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 2, + "regex": "", + "sort": 5, + "type": "query" + }, + { + "allValue": ".*", + "current": { + "text": "All", + "value": "$__all" + }, + "definition": "label_values(kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",namespace=~\"$NameSpace\"},container)", + "includeAll": true, + "label": "Microservice (Container Name)", + "name": "Container", + "options": [], + "query": { + "query": "label_values(kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",namespace=~\"$NameSpace\"},container)", + "refId": "StandardVariableQuery" + }, + "refresh": 2, + "regex": "", + "sort": 5, + "type": "query" + }, + { + "allValue": ".*", + "current": { + "text": "All", + "value": "$__all" + }, + "definition": "label_values(kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",namespace=~\"$NameSpace\",container=~\"$Container\"},pod)", + "includeAll": true, + "label": "Pod", + "name": "Pod", + "options": [], + "query": { + "query": "label_values(kube_pod_container_info{origin_prometheus=~\"$origin_prometheus\",namespace=~\"$NameSpace\",container=~\"$Container\"},pod)", + "refId": "StandardVariableQuery" + }, + "refresh": 2, + "regex": "", + "sort": 5, + "type": "query" + } + ] + }, + "time": { + "from": "now-30m", + "to": "now" + }, + "timepicker": { + "refresh_intervals": ["30s", "1m", "5m", "15m", "30m", "1h"] + }, + "timezone": "", + "title": "K8S Dashboard EN 20250125", + "uid": "15661", + "version": 3 +} diff --git a/rootfs/usr/share/grafana/dashboards/kubernetes_apiserver.json b/rootfs/usr/share/grafana/dashboards/kubernetes_apiserver.json new file mode 100644 index 0000000..9862cf6 --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/kubernetes_apiserver.json @@ -0,0 +1,639 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": "-- Grafana --", + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "editable": false, + "gnetId": 12006, + "graphTooltip": 0, + "id": null, + "iteration": 1585691077160, + "links": [], + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 0 + }, + "hiddenSeries": false, + "id": 8, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(apiserver_request_total[5m])) by (verb)", + "legendFormat": "{{verb}}", + "refId": "A" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "apiserver request rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "short", + "label": "req/s", + "logBase": 1, + "max": null, + "min": "0", + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 0 + }, + "hiddenSeries": false, + "id": 10, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.95, sum(rate(workqueue_queue_duration_seconds_bucket[5m])) by (le))", + "legendFormat": "p95", + "refId": "A" + }, + { + "expr": "histogram_quantile(0.90, sum(rate(workqueue_queue_duration_seconds_bucket[5m])) by (le))", + "legendFormat": "p90", + "refId": "B" + }, + { + "expr": "histogram_quantile(0.5, sum(rate(workqueue_queue_duration_seconds_bucket[5m])) by (le))", + "legendFormat": "p50", + "refId": "C" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "workqueue service time", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "s", + "label": "service time", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "description": "95th percentile request latency for Kubernetes apiserver component", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 8 + }, + "hiddenSeries": false, + "id": 2, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.95, sum(rate(apiserver_request_duration_seconds_bucket{verb!~\"CONNECT|WATCH\"}[5m])) by (le))", + "legendFormat": "p95", + "refId": "A" + }, + { + "expr": "histogram_quantile(0.90, sum(rate(apiserver_request_duration_seconds_bucket{verb!~\"CONNECT|WATCH\"}[5m])) by (le))", + "legendFormat": "p90", + "refId": "B" + }, + { + "expr": "histogram_quantile(0.5, sum(rate(apiserver_request_duration_seconds_bucket{verb!~\"CONNECT|WATCH\"}[5m])) by (le))", + "legendFormat": "p50 ", + "refId": "C" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "apiserver request latency", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "decimals": null, + "format": "s", + "label": "latency", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 8 + }, + "hiddenSeries": false, + "id": 12, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.95, sum(rate(workqueue_work_duration_seconds_bucket[5m])) by (le))", + "legendFormat": "p95", + "refId": "A" + }, + { + "expr": "histogram_quantile(0.9, sum(rate(workqueue_work_duration_seconds_bucket[5m])) by (le))", + "legendFormat": "p90", + "refId": "B" + }, + { + "expr": "histogram_quantile(0.5, sum(rate(workqueue_work_duration_seconds_bucket[5m])) by (le))", + "legendFormat": "p50", + "refId": "C" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "workqueue processing time", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "s", + "label": "processing time", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 16 + }, + "hiddenSeries": false, + "id": 6, + "interval": "", + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(etcd_helper_cache_hit_total[5m]))/sum((rate(etcd_helper_cache_hit_total[5m])+rate(etcd_helper_cache_miss_total[5m])))", + "legendFormat": "ratio", + "refId": "A" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "etcd helper cache hit ratio", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "decimals": null, + "format": "percentunit", + "label": "", + "logBase": 1, + "max": null, + "min": "0", + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "description": "etcd request latency (p95)", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 17 + }, + "hiddenSeries": false, + "id": 4, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "dataLinks": [] + }, + "percentage": false, + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.95, sum(rate(etcd_request_duration_seconds_bucket[5m])) by (le))", + "legendFormat": "p95", + "refId": "A" + }, + { + "expr": "histogram_quantile(0.90, sum(rate(etcd_request_duration_seconds_bucket[5m])) by (le))", + "legendFormat": "p90", + "refId": "B" + }, + { + "expr": "histogram_quantile(0.5, sum(rate(etcd_request_duration_seconds_bucket[5m])) by (le))", + "legendFormat": "p50", + "refId": "C" + } + ], + "thresholds": [], + "timeFrom": null, + "timeRegions": [], + "timeShift": null, + "title": "etcd request latency", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "decimals": null, + "format": "s", + "label": "latency", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + } + ], + "refresh": false, + "schemaVersion": 22, + "style": "dark", + "tags": [], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "includeAll": false, + "label": "Data source", + "multi": false, + "name": "datasource", + "options": [], + "query": "prometheus", + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "type": "datasource" + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ] + }, + "timezone": "", + "title": "Kubernetes apiserver", + "uid": "R6abPf9Zz", + "version": 3, + "description": "Latency and cache hit graphs for Kubernetes apiserver. Useful for observing performance of Amazon EKS clusters." +} diff --git a/rootfs/usr/share/grafana/dashboards/kubernetes_cluster.json b/rootfs/usr/share/grafana/dashboards/kubernetes_cluster.json new file mode 100644 index 0000000..8a0b173 --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/kubernetes_cluster.json @@ -0,0 +1,2010 @@ +{ + "id": null, + "title": "Kubernetes cluster monitoring (via Prometheus)", + "description": "Monitors Kubernetes cluster using Prometheus. Shows overall cluster CPU / Memory / Filesystem usage as well as individual pod, containers, systemd services statistics. Uses cAdvisor metrics only.", + "tags": ["kubernetes"], + "style": "dark", + "timezone": "browser", + "editable": false, + "hideControls": false, + "sharedCrosshair": false, + "rows": [ + { + "collapse": false, + "editable": false, + "height": "200px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 2, + "editable": false, + "error": false, + "fill": 1, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)", + "thresholdLine": false + }, + "height": "200px", + "id": 32, + "isNew": true, + "legend": { + "alignAsTable": false, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": false, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum (rate (container_network_receive_bytes_total{kubernetes_io_hostname=~\"^$Node$\"}[1m]))", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "Received", + "metric": "network", + "refId": "A", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{kubernetes_io_hostname=~\"^$Node$\"}[1m]))", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "Sent", + "metric": "network", + "refId": "B", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "Network I/O pressure", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "cumulative" + }, + "transparent": false, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "title": "Network I/O pressure" + }, + { + "collapse": false, + "editable": false, + "height": "250px", + "panels": [ + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": true, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "editable": false, + "error": false, + "format": "percent", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": true, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "180px", + "id": 4, + "interval": null, + "isNew": true, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 4, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{id=\"/\",kubernetes_io_hostname=~\"^$Node$\"}) / sum (machine_memory_bytes{kubernetes_io_hostname=~\"^$Node$\"}) * 100", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "65, 90", + "title": "Cluster memory usage", + "transparent": false, + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": true, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "decimals": 2, + "editable": false, + "error": false, + "format": "percent", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": true, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "180px", + "id": 6, + "interval": null, + "isNew": true, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 4, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{id=\"/\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) / sum (machine_cpu_cores{kubernetes_io_hostname=~\"^$Node$\"}) * 100", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "65, 90", + "title": "Cluster CPU usage (1m avg)", + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": true, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "decimals": 2, + "editable": false, + "error": false, + "format": "percent", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": true, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "180px", + "id": 7, + "interval": null, + "isNew": true, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 4, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "targets": [ + { + "expr": "sum (container_fs_usage_bytes{device=~\"^/dev/.*$\",id=\"/\",kubernetes_io_hostname=~\"^$Node$\"}) / sum (container_fs_limit_bytes{device=~\"^/dev/.*$\",id=\"/\",kubernetes_io_hostname=~\"^$Node$\"}) * 100", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "", + "metric": "", + "refId": "A", + "step": 10 + } + ], + "thresholds": "65, 90", + "title": "Cluster filesystem usage", + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "decimals": 2, + "editable": false, + "error": false, + "format": "bytes", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 9, + "interval": null, + "isNew": true, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "20%", + "prefix": "", + "prefixFontSize": "20%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{id=\"/\",kubernetes_io_hostname=~\"^$Node$\"})", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Used", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "decimals": 2, + "editable": false, + "error": false, + "format": "bytes", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 10, + "interval": null, + "isNew": true, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "targets": [ + { + "expr": "sum (machine_memory_bytes{kubernetes_io_hostname=~\"^$Node$\"})", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Total", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "decimals": 2, + "editable": false, + "error": false, + "format": "none", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 11, + "interval": null, + "isNew": true, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": " cores", + "postfixFontSize": "30%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{id=\"/\",kubernetes_io_hostname=~\"^$Node$\"}[1m]))", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Used", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "decimals": 2, + "editable": false, + "error": false, + "format": "none", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 12, + "interval": null, + "isNew": true, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": " cores", + "postfixFontSize": "30%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "targets": [ + { + "expr": "sum (machine_cpu_cores{kubernetes_io_hostname=~\"^$Node$\"})", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Total", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "decimals": 2, + "editable": false, + "error": false, + "format": "bytes", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 13, + "interval": null, + "isNew": true, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "targets": [ + { + "expr": "sum (container_fs_usage_bytes{device=~\"^/dev/.*$\",id=\"/\",kubernetes_io_hostname=~\"^$Node$\"})", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Used", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "decimals": 2, + "editable": false, + "error": false, + "format": "bytes", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "height": "1px", + "id": 14, + "interval": null, + "isNew": true, + "links": [], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "targets": [ + { + "expr": "sum (container_fs_limit_bytes{device=~\"^/dev/.*$\",id=\"/\",kubernetes_io_hostname=~\"^$Node$\"})", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 10 + } + ], + "thresholds": "", + "title": "Total", + "type": "singlestat", + "valueFontSize": "50%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + } + ], + "showTitle": false, + "title": "Total usage" + }, + { + "collapse": false, + "editable": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 3, + "editable": false, + "error": false, + "fill": 0, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)" + }, + "height": "", + "id": 17, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",name=~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (pod_name)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ pod_name }}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "Pods CPU usage (1m avg)", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "transparent": false, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "none", + "label": "cores", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "showTitle": false, + "title": "Pods CPU usage" + }, + { + "collapse": true, + "editable": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 3, + "editable": false, + "error": false, + "fill": 0, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)" + }, + "height": "", + "id": 23, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{systemd_service_name!=\"\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (systemd_service_name)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ systemd_service_name }}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "System services CPU usage (1m avg)", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "none", + "label": "cores", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "title": "System services CPU usage" + }, + { + "collapse": true, + "editable": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 3, + "editable": false, + "error": false, + "fill": 0, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)" + }, + "height": "", + "id": 24, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": false, + "hideZero": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",name=~\"^k8s_.*\",container_name!=\"POD\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (container_name, pod_name)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "pod: {{ pod_name }} | {{ container_name }}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + }, + { + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",name!~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (kubernetes_io_hostname, name, image)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})", + "metric": "container_cpu", + "refId": "B", + "step": 10 + }, + { + "expr": "sum (rate (container_cpu_usage_seconds_total{rkt_container_name!=\"\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (kubernetes_io_hostname, rkt_container_name)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}", + "metric": "container_cpu", + "refId": "C", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "Containers CPU usage (1m avg)", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "none", + "label": "cores", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "title": "Containers CPU usage" + }, + { + "collapse": true, + "editable": false, + "height": "500px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 3, + "editable": false, + "error": false, + "fill": 0, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)" + }, + "id": 20, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (rate (container_cpu_usage_seconds_total{id!=\"/\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (id)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ id }}", + "metric": "container_cpu", + "refId": "A", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "All processes CPU usage (1m avg)", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "none", + "label": "cores", + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "showTitle": false, + "title": "All processes CPU usage" + }, + { + "collapse": false, + "editable": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 2, + "editable": false, + "error": false, + "fill": 0, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)" + }, + "id": 25, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{image!=\"\",name=~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}) by (pod_name)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ pod_name }}", + "metric": "container_memory_usage:sort_desc", + "refId": "A", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "Pods memory usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "title": "Pods memory usage" + }, + { + "collapse": true, + "editable": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 2, + "editable": false, + "error": false, + "fill": 0, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)" + }, + "id": 26, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{systemd_service_name!=\"\",kubernetes_io_hostname=~\"^$Node$\"}) by (systemd_service_name)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ systemd_service_name }}", + "metric": "container_memory_usage:sort_desc", + "refId": "A", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "System services memory usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "title": "System services memory usage" + }, + { + "collapse": true, + "editable": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 2, + "editable": false, + "error": false, + "fill": 0, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)" + }, + "id": 27, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{image!=\"\",name=~\"^k8s_.*\",container_name!=\"POD\",kubernetes_io_hostname=~\"^$Node$\"}) by (container_name, pod_name)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "pod: {{ pod_name }} | {{ container_name }}", + "metric": "container_memory_usage:sort_desc", + "refId": "A", + "step": 10 + }, + { + "expr": "sum (container_memory_working_set_bytes{image!=\"\",name!~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}) by (kubernetes_io_hostname, name, image)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})", + "metric": "container_memory_usage:sort_desc", + "refId": "B", + "step": 10 + }, + { + "expr": "sum (container_memory_working_set_bytes{rkt_container_name!=\"\",kubernetes_io_hostname=~\"^$Node$\"}) by (kubernetes_io_hostname, rkt_container_name)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}", + "metric": "container_memory_usage:sort_desc", + "refId": "C", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "Containers memory usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "title": "Containers memory usage" + }, + { + "collapse": true, + "editable": false, + "height": "500px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 2, + "editable": false, + "error": false, + "fill": 0, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)" + }, + "id": 28, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": true, + "targets": [ + { + "expr": "sum (container_memory_working_set_bytes{id!=\"/\",kubernetes_io_hostname=~\"^$Node$\"}) by (id)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ id }}", + "metric": "container_memory_usage:sort_desc", + "refId": "A", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "All processes memory usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "title": "All processes memory usage" + }, + { + "collapse": false, + "editable": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 2, + "editable": false, + "error": false, + "fill": 1, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)" + }, + "id": 16, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum (rate (container_network_receive_bytes_total{image!=\"\",name=~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (pod_name)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "-> {{ pod_name }}", + "metric": "network", + "refId": "A", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{image!=\"\",name=~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (pod_name)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "<- {{ pod_name }}", + "metric": "network", + "refId": "B", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "Pods network I/O (1m avg)", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "title": "Pods network I/O" + }, + { + "collapse": true, + "editable": false, + "height": "250px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 2, + "editable": false, + "error": false, + "fill": 1, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)" + }, + "id": 30, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum (rate (container_network_receive_bytes_total{image!=\"\",name=~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (container_name, pod_name)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "-> pod: {{ pod_name }} | {{ container_name }}", + "metric": "network", + "refId": "B", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{image!=\"\",name=~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (container_name, pod_name)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "<- pod: {{ pod_name }} | {{ container_name }}", + "metric": "network", + "refId": "D", + "step": 10 + }, + { + "expr": "sum (rate (container_network_receive_bytes_total{image!=\"\",name!~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (kubernetes_io_hostname, name, image)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "-> docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})", + "metric": "network", + "refId": "A", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{image!=\"\",name!~\"^k8s_.*\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (kubernetes_io_hostname, name, image)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "<- docker: {{ kubernetes_io_hostname }} | {{ image }} ({{ name }})", + "metric": "network", + "refId": "C", + "step": 10 + }, + { + "expr": "sum (rate (container_network_transmit_bytes_total{rkt_container_name!=\"\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (kubernetes_io_hostname, rkt_container_name)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "-> rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}", + "metric": "network", + "refId": "E", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{rkt_container_name!=\"\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (kubernetes_io_hostname, rkt_container_name)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "<- rkt: {{ kubernetes_io_hostname }} | {{ rkt_container_name }}", + "metric": "network", + "refId": "F", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "Containers network I/O (1m avg)", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "title": "Containers network I/O" + }, + { + "collapse": true, + "editable": false, + "height": "500px", + "panels": [ + { + "aliasColors": {}, + "bars": false, + "decimals": 2, + "editable": false, + "error": false, + "fill": 1, + "grid": { + "threshold1": null, + "threshold1Color": "rgba(216, 200, 27, 0.27)", + "threshold2": null, + "threshold2Color": "rgba(234, 112, 112, 0.22)" + }, + "id": 29, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum (rate (container_network_receive_bytes_total{id!=\"/\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (id)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "-> {{ id }}", + "metric": "network", + "refId": "A", + "step": 10 + }, + { + "expr": "- sum (rate (container_network_transmit_bytes_total{id!=\"/\",kubernetes_io_hostname=~\"^$Node$\"}[1m])) by (id)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "<- {{ id }}", + "metric": "network", + "refId": "B", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "All processes network I/O (1m avg)", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "show": true + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "title": "All processes network I/O" + } + ], + "time": { + "from": "now-5m", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": ["5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d"] + }, + "templating": { + "list": [ + { + "allValue": ".*", + "current": {}, + "hide": 0, + "includeAll": true, + "multi": false, + "name": "Node", + "options": [], + "query": "label_values(kubernetes_io_hostname)", + "refresh": 1, + "type": "query" + } + ] + }, + "annotations": { + "list": [] + }, + "refresh": "10s", + "schemaVersion": 12, + "version": 13, + "uid": "315", + "links": [], + "gnetId": 1621 +} diff --git a/rootfs/usr/share/grafana/dashboards/kubernetes_deployment_metrics.json b/rootfs/usr/share/grafana/dashboards/kubernetes_deployment_metrics.json new file mode 100644 index 0000000..a9062b7 --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/kubernetes_deployment_metrics.json @@ -0,0 +1,1324 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "target": { + "limit": 100, + "matchAny": false, + "tags": [], + "type": "dashboard" + }, + "type": "dashboard" + } + ] + }, + "description": "Monitors Kubernetes deployments in cluster using Prometheus. Shows overall cluster CPU / Memory / Network of deployments, replicas in each deployment. Uses Kube state metrics and cAdvisor metrics ", + "editable": false, + "fiscalYearStartMonth": 0, + "gnetId": 741, + "graphTooltip": 0, + "id": null, + "links": [], + "liveNow": false, + "panels": [ + { + "datasource": null, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 65 + }, + { + "color": "rgba(245, 54, 54, 0.9)", + "value": 90 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 8, + "x": 0, + "y": 0 + }, + "id": 4, + "links": [], + "maxDataPoints": 100, + "options": { + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showThresholdLabels": false, + "showThresholdMarkers": true + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "datasource": null, + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{pod=~\"^$Deployment.*$\",pod=~\"^$Statefulset.*$\",namespace=\"$Namespace\"}) / sum (machine_memory_bytes{kubernetes_io_hostname=~\"^$Node$\"}) * 100", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "", + "range": true, + "refId": "A", + "step": 10 + } + ], + "title": "Deployment memory usage", + "type": "gauge" + }, + { + "datasource": null, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "decimals": 2, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 65 + }, + { + "color": "rgba(245, 54, 54, 0.9)", + "value": 90 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 8, + "x": 8, + "y": 0 + }, + "id": 6, + "links": [], + "maxDataPoints": 100, + "options": { + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showThresholdLabels": false, + "showThresholdMarkers": true + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "datasource": null, + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{pod=~\"^$Deployment.*$\",pod=~\"^$Statefulset.*$\",namespace=\"$Namespace\"}[1m])) / sum (machine_cpu_cores{kubernetes_io_hostname=~\"^$Node$\"}) * 100", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "", + "range": true, + "refId": "A", + "step": 10 + } + ], + "title": "Deployment CPU usage", + "type": "gauge" + }, + { + "datasource": null, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 8, + "x": 16, + "y": 0 + }, + "id": 37, + "links": [], + "maxDataPoints": 100, + "options": { + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "showThresholdLabels": false, + "showThresholdMarkers": true + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "datasource": null, + "editorMode": "code", + "expr": "sum(kube_deployment_status_replicas_available{deployment=~\"^$Deployment$\",pod=~\"^$Statefulset.*$\",namespace=\"$Namespace\"}) / sum(kube_deployment_status_replicas{deployment=~\"^$Deployment$\",pod=~\"^$Statefulset.*$\",namespace=\"$Namespace\"}) * 100", + "intervalFactor": 2, + "range": true, + "refId": "A", + "step": 40 + } + ], + "title": "Replicas", + "type": "gauge" + }, + { + "datasource": null, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 0, + "y": 5 + }, + "id": 38, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "datasource": null, + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{namespace=\"$Namespace\",pod=~\"^$Deployment.*$\",pod=~\"^$Statefulset.*$\"})", + "interval": "", + "intervalFactor": 2, + "legendFormat": "", + "range": true, + "refId": "A", + "step": 40 + } + ], + "title": "Used", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 4, + "y": 5 + }, + "id": 39, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "datasource": null, + "expr": "sum (container_memory_working_set_bytes{kubernetes_io_hostname=~\"^$Node.*$\"})", + "intervalFactor": 2, + "refId": "A", + "step": 40 + } + ], + "title": "Total", + "type": "stat" + }, + { + "datasource": null, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 8, + "y": 5 + }, + "id": 40, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "datasource": null, + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{pod=~\"^$Deployment.*$\",pod=~\"^$Statefulset.*$\",namespace=\"$Namespace\"}[1m]))", + "interval": "", + "intervalFactor": 2, + "legendFormat": "", + "range": true, + "refId": "A", + "step": 40 + } + ], + "title": "Used", + "type": "stat" + }, + { + "datasource": null, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 12, + "y": 5 + }, + "id": 41, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "datasource": null, + "expr": "sum (machine_cpu_cores{kubernetes_io_hostname=~\"^$Node$\"})", + "intervalFactor": 2, + "refId": "A", + "step": 40 + } + ], + "title": "Total", + "type": "stat" + }, + { + "datasource": null, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 16, + "y": 5 + }, + "id": 42, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "datasource": null, + "editorMode": "code", + "expr": "sum(kube_deployment_status_replicas_available{namespace=\"$Namespace\",deployment=~\"^$Deployment$\",pod=~\"^$Statefulset.*$\"})", + "intervalFactor": 2, + "range": true, + "refId": "A", + "step": 40 + } + ], + "title": "Available", + "type": "stat" + }, + { + "datasource": null, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 20, + "y": 5 + }, + "id": 43, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "datasource": null, + "editorMode": "code", + "expr": "sum(kube_deployment_status_replicas{namespace=\"$Namespace\",deployment=~\"^$Deployment$\",pod=~\"^$Statefulset.*$\"})", + "intervalFactor": 2, + "range": true, + "refId": "A", + "step": 40 + } + ], + "title": "Total", + "type": "stat" + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": null, + "decimals": 3, + "editable": false, + "error": false, + "fill": 0, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 8 + }, + "height": "", + "hiddenSeries": false, + "id": 17, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": null, + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{image!=\"\",pod!=\"\",pod=~\"^$Deployment.*$\",pod=~\"^$Statefulset.*$\",namespace=\"$Namespace\"}[1m])) by (pod)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ pod }}", + "metric": "container_cpu", + "range": true, + "refId": "A", + "step": 10 + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Pod CPU usage", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:677", + "format": "none", + "label": "cores", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:678", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": null, + "decimals": 3, + "editable": false, + "error": false, + "fill": 0, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 15 + }, + "height": "", + "hiddenSeries": false, + "id": 24, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": false, + "hideZero": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": null, + "editorMode": "code", + "expr": "sum (rate (container_cpu_usage_seconds_total{pod!=\"\",pod!=\"POD\",pod=~\"^$Deployment.*$\",pod=~\"^$Statefulset.*$\",namespace=\"$Namespace\"}[1m])) by (pod, container)", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "container: {{ pod }}:{{ container }} ", + "metric": "container_cpu", + "range": true, + "refId": "A", + "step": 10 + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Containers CPU usage", + "tooltip": { + "msResolution": true, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:189", + "format": "none", + "label": "cores", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:190", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": null, + "decimals": 2, + "editable": false, + "error": false, + "fill": 0, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 22 + }, + "hiddenSeries": false, + "id": 25, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": null, + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",pod=~\"^$Deployment.*$\",pod=~\"^$Statefulset.*$\",namespace=\"$Namespace\"}) by (pod)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ pod }}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "A", + "step": 10 + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Pod memory usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:920", + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:921", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": {}, + "decimals": 2, + "editable": false, + "error": false, + "fill": 0, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 29 + }, + "hiddenSeries": false, + "id": 44, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": {}, + "editorMode": "code", + "expr": "sum (container_memory_working_set_bytes{image!=\"\",pod!=\"\",pod=~\"^$Deployment.*$\",pod=~\"^$Statefulset.*$\",namespace=\"$Namespace\"}) by (pod, container)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{ pod }}:{{ container }}", + "metric": "container_memory_usage:sort_desc", + "range": true, + "refId": "A", + "step": 10 + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Container memory usage", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:920", + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:921", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": null, + "decimals": 2, + "editable": false, + "error": false, + "fill": 1, + "fillGradient": 0, + "grid": {}, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 36 + }, + "hiddenSeries": false, + "id": 16, + "isNew": true, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": 200, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": null, + "editorMode": "code", + "expr": "sum (rate (container_network_receive_bytes_total{image!=\"\",id!=\"\",pod=~\"^$Deployment.*$\",pod=~\"^$Statefulset.*$\",namespace=\"${Namespace}\"}[2m])) by (pod)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "-> {{ pod }}", + "metric": "network", + "range": true, + "refId": "A", + "step": 10 + }, + { + "datasource": null, + "editorMode": "code", + "expr": "- sum (rate (container_network_transmit_bytes_total{image!=\"\",id!=\"\",pod=~\"^$Deployment.*$\",pod=~\"^$Statefulset.*$\",namespace=\"${Namespace}\"}[2m])) by (pod)", + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "<- {{ pod }}", + "metric": "network", + "range": true, + "refId": "B", + "step": 10 + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Pod network I/O", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 2, + "value_type": "cumulative" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:514", + "format": "Bps", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:515", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + } + ], + "refresh": "10s", + "schemaVersion": 37, + "style": "dark", + "tags": ["kubernetes"], + "templating": { + "list": [ + { + "current": { + "selected": true, + "text": "", + "value": "" + }, + "datasource": null, + "definition": "label_values(namespace)", + "hide": 0, + "includeAll": false, + "multi": false, + "name": "Namespace", + "options": [], + "query": { + "query": "label_values(namespace)", + "refId": "StandardVariableQuery" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "type": "query" + }, + { + "allValue": ".*", + "current": { + "selected": true, + "text": ["All"], + "value": ["$__all"] + }, + "datasource": {}, + "definition": "label_values({namespace=\"$Namespace\"},deployment)", + "hide": 0, + "includeAll": true, + "multi": true, + "name": "Deployment", + "options": [], + "query": { + "query": "label_values({namespace=\"$Namespace\"},deployment)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "tagValuesQuery": "", + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": ".*", + "current": { + "selected": true, + "text": ["All"], + "value": ["$__all"] + }, + "definition": "label_values({namespace=\"$Namespace\"},statefulset)", + "hide": 0, + "includeAll": true, + "multi": true, + "name": "Statefulset", + "options": [], + "query": { + "query": "label_values({namespace=\"$Namespace\"},statefulset)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "type": "query" + }, + { + "allValue": ".*", + "current": { + "selected": true, + "text": "All", + "value": "$__all" + }, + "datasource": {}, + "definition": "label_values(node)", + "hide": 0, + "includeAll": true, + "multi": false, + "name": "Node", + "options": [], + "query": { + "query": "label_values(node)", + "refId": "StandardVariableQuery" + }, + "refresh": 1, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "tagValuesQuery": "", + "tagsQuery": "", + "type": "query", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": ["5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d"] + }, + "timezone": "browser", + "title": "Kubernetes Deployment metrics", + "uid": "wdtm7Mc7z123", + "version": 1, + "weekStart": "" +} diff --git a/rootfs/usr/share/grafana/dashboards/kubernetes_metrics.json b/rootfs/usr/share/grafana/dashboards/kubernetes_metrics.json new file mode 100644 index 0000000..5e07a52 --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/kubernetes_metrics.json @@ -0,0 +1,3390 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "description": "Summary metrics about kube-state-metrics v2 version(https://github.com/kubernetes/kube-state-metrics); Referenced 6417", + "editable": false, + "fiscalYearStartMonth": 0, + "graphTooltip": 1, + "id": null, + "links": [ + { + "asDropdown": true, + "icon": "external link", + "includeVars": true, + "keepTime": false, + "tags": ["kubernetes-app"], + "title": "Dashboards", + "type": "dashboards" + } + ], + "panels": [ + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 58, + "panels": [], + "title": "Cluster", + "type": "row" + }, + { + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "#299c46" + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 80 + }, + { + "color": "#d44a3a", + "value": 90 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 6, + "x": 0, + "y": 1 + }, + "id": 4, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_info{cluster=~\"$cluster\",node=~\"$node\"}) / sum(kube_node_status_allocatable{cluster=~\"$cluster\",resource=\"pods\",node=~\"$node\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Cluster Pod Requested", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "#299c46" + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 80 + }, + { + "color": "#d44a3a", + "value": 90 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 6, + "x": 6, + "y": 1 + }, + "id": 5, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_container_resource_requests{cluster=~\"$cluster\",resource=\"cpu\",node=~\"$node\"})/ sum(kube_node_status_allocatable{node=~\"$node\",cluster=~\"$cluster\",resource=\"cpu\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Cluster CPU Requested", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "#299c46" + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 80 + }, + { + "color": "#d44a3a", + "value": 90 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 6, + "x": 12, + "y": 1 + }, + "id": 6, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_container_resource_requests{cluster=~\"$cluster\",resource=\"memory\",node=~\"$node\"}) / sum(kube_node_status_allocatable{node=~\"$node\",cluster=~\"$cluster\",resource=\"memory\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Cluster Memory Requested", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "pods", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 6, + "x": 0, + "y": 5 + }, + "id": 9, + "options": { + "alertThreshold": true, + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_node_status_allocatable{cluster=~\"$cluster\",resource=\"pods\",node=~\"$node\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "allocatable", + "refId": "A" + }, + { + "expr": "sum(kube_pod_info{node=~\"$node\",cluster=~\"$cluster\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "requested", + "refId": "C" + } + ], + "title": "Cluster Pod Capacity", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "cores", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 6, + "x": 6, + "y": 5 + }, + "id": 10, + "options": { + "alertThreshold": true, + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_node_status_capacity{node=~\"$node\",cluster=~\"$cluster\",resource=\"cpu\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "allocatable", + "refId": "A" + }, + { + "expr": "sum(kube_node_status_allocatable{node=~\"$node\",cluster=~\"$cluster\",resource=\"cpu\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "capacity", + "refId": "B" + }, + { + "expr": "sum(kube_pod_container_resource_requests{cluster=~\"$cluster\",resource=\"cpu\",node=~\"$node\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "requested", + "refId": "C" + }, + { + "expr": "sum(kube_pod_container_resource_limits{cluster=~\"$cluster\",resource=\"cpu\",node=~\"$node\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "limited", + "refId": "D" + } + ], + "title": "Cluster CPU Capacity", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bits" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 6, + "x": 12, + "y": 5 + }, + "id": 11, + "options": { + "alertThreshold": true, + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_node_status_allocatable{node=~\"$node\",cluster=~\"$cluster\",resource=\"memory\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "allocatable", + "refId": "A" + }, + { + "expr": "sum(kube_node_status_capacity{node=~\"$node\",cluster=~\"$cluster\",resource=\"memory\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "capacity", + "refId": "B" + }, + { + "expr": "sum(kube_pod_container_resource_requests{cluster=~\"$cluster\",resource=\"memory\",node=~\"$node\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "requested", + "refId": "C" + }, + { + "expr": "sum(kube_pod_container_resource_limits{cluster=~\"$cluster\",resource=\"memory\",node=~\"$node\"})", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "limited", + "refId": "D" + } + ], + "title": "Cluster Mem Capacity", + "type": "timeseries" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 10 + }, + "id": 22, + "panels": [ + { + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 0, + "y": 20 + }, + "id": 24, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_node_info{cluster=~\"$cluster\"})", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Number Of Nodes", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "#299c46" + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 1 + }, + { + "color": "#d44a3a" + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 4, + "x": 4, + "y": 20 + }, + "id": 26, + "maxDataPoints": 100, + "options": { + "colorMode": "background", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_node_spec_unschedulable{cluster=~\"$cluster\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Nodes Unavailable", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 8, + "x": 8, + "y": 20 + }, + "id": 78, + "options": { + "alertThreshold": true, + "legend": { + "calcs": [], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "kube_node_status_condition{cluster=~\"$cluster\",condition=\"Ready\",status=\"false\"}==1", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{node}}", + "refId": "A" + } + ], + "title": "Node NotReady", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 8, + "x": 16, + "y": 20 + }, + "id": 51, + "options": { + "alertThreshold": true, + "legend": { + "calcs": [], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "kube_node_status_condition{condition=\"DiskPressure\",cluster=~\"$cluster\",status=\"true\"}==1", + "format": "time_series", + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{node}} {{condition}}", + "refId": "A" + }, + { + "expr": "kube_node_status_condition{condition=\"MemoryPressure\",node=~\"$node\",cluster=~\"$cluster\",status=\"true\"}==1", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{node}} {{condition}}", + "refId": "B" + }, + { + "expr": "kube_node_status_condition{condition=~\"PIDPressure\",node=~\"$node\",cluster=~\"$cluster\",status=\"true\"}==1", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{node}} {{condition}}", + "refId": "C" + } + ], + "title": "Node Pressure", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 23 + }, + "id": 88, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max", "min"], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_node_info{cluster=~\"$cluster\"})", + "interval": "", + "legendFormat": "", + "refId": "A" + } + ], + "title": "node number", + "type": "timeseries" + } + ], + "title": "Node", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 11 + }, + "id": 14, + "panels": [ + { + "fieldConfig": { + "defaults": { + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "inspect": false + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Time" + }, + "properties": [ + { + "id": "displayName", + "value": "Time" + }, + { + "id": "unit", + "value": "time: YYYY-MM-DD HH:mm:ss" + }, + { + "id": "custom.align" + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Metric" + }, + "properties": [ + { + "id": "unit", + "value": "none" + }, + { + "id": "custom.cellOptions", + "value": { + "type": "color-background" + } + }, + { + "id": "custom.align" + }, + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(245, 54, 54, 0.9)" + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 0 + }, + { + "color": "rgba(50, 172, 45, 0.97)", + "value": 0 + }, + {} + ] + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Value" + }, + "properties": [ + { + "id": "unit", + "value": "none" + }, + { + "id": "custom.cellOptions", + "value": { + "type": "color-background" + } + }, + { + "id": "custom.align" + }, + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(245, 54, 54, 0.9)" + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 0 + }, + { + "color": "rgba(50, 172, 45, 0.97)", + "value": 1 + } + ] + } + } + ] + } + ] + }, + "gridPos": { + "h": 5, + "w": 6, + "x": 0, + "y": 218 + }, + "id": 16, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "kube_deployment_status_replicas{namespace=~\"$namespace\",cluster=~\"$cluster\"}", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{ deployment }}", + "refId": "A" + } + ], + "title": "Deployment Replicas - Up To Date", + "type": "table" + }, + { + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 6, + "x": 6, + "y": 218 + }, + "id": 18, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_deployment_status_replicas{namespace=~\"$namespace\",cluster=~\"$cluster\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Deployment Replicas", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 6, + "x": 12, + "y": 218 + }, + "id": 19, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_deployment_status_replicas_updated{namespace=~\"$namespace\",cluster=~\"$cluster\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Deployment Replicas - Updated", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 6, + "x": 18, + "y": 218 + }, + "id": 20, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["mean"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_deployment_status_replicas_unavailable{namespace=~\"$namespace\",cluster=~\"$cluster\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Deployment Replicas - Unavailable", + "type": "stat" + } + ], + "title": "Deployments", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 12 + }, + "id": 71, + "panels": [ + { + "fieldConfig": { + "defaults": { + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "inspect": false + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Time" + }, + "properties": [ + { + "id": "displayName", + "value": "Time" + }, + { + "id": "unit", + "value": "time: YYYY-MM-DD HH:mm:ss" + }, + { + "id": "custom.align" + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Metric" + }, + "properties": [ + { + "id": "unit", + "value": "none" + }, + { + "id": "custom.cellOptions", + "value": { + "type": "color-background" + } + }, + { + "id": "custom.align" + }, + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(245, 54, 54, 0.9)" + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 0 + }, + { + "color": "rgba(50, 172, 45, 0.97)", + "value": 0 + }, + {} + ] + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Value" + }, + "properties": [ + { + "id": "unit", + "value": "none" + }, + { + "id": "custom.cellOptions", + "value": { + "type": "color-background" + } + }, + { + "id": "custom.align" + }, + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(245, 54, 54, 0.9)" + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 0 + }, + { + "color": "rgba(50, 172, 45, 0.97)", + "value": 1 + } + ] + } + } + ] + } + ] + }, + "gridPos": { + "h": 5, + "w": 6, + "x": 0, + "y": 22 + }, + "id": 75, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "kube_statefulset_status_replicas_ready{namespace=~\"$namespace\",cluster=~\"$cluster\"}", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{ statefulset}}", + "refId": "A" + } + ], + "title": "Statefulset Replicas - Up To Date", + "type": "table" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 12, + "x": 6, + "y": 22 + }, + "id": 73, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "kube_statefulset_status_replicas_ready{namespace=~\"$namespace\",cluster=~\"$cluster\"}/kube_statefulset_status_replicas{namespace=~\"$namespace\",cluster=~\"$cluster\"}*100", + "format": "time_series", + "instant": false, + "intervalFactor": 1, + "legendFormat": "{{statefulset}}", + "refId": "A" + } + ], + "title": "Statefulset replicas", + "type": "timeseries" + } + ], + "title": "Statefuleset", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 13 + }, + "id": 28, + "panels": [ + { + "fieldConfig": { + "defaults": { + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "inspect": false + }, + "decimals": 2, + "displayName": "", + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "#56A64B" + }, + { + "color": "#73BF69" + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "Time" + }, + "properties": [ + { + "id": "displayName", + "value": "Time" + }, + { + "id": "unit", + "value": "short" + }, + { + "id": "decimals", + "value": 2 + }, + { + "id": "custom.hidden", + "value": true + }, + { + "id": "custom.align" + } + ] + } + ] + }, + "gridPos": { + "h": 9, + "w": 6, + "x": 0, + "y": 23 + }, + "id": 68, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "changes(kube_pod_container_status_restarts_total{namespace=~\"$namespace\",cluster=~\"$cluster\"}[30m])>1", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A" + } + ], + "title": "Pods restart in 30m", + "type": "table" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 0, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 6, + "x": 6, + "y": 23 + }, + "id": 77, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "kube_pod_status_reason{cluster=~\"$cluster\",namespace=~\"$namespace\", reason=\"Evicted\"}==1", + "format": "time_series", + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A" + } + ], + "title": "Evicted", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "fixedColor": "#629e51", + "mode": "fixed" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 12, + "x": 12, + "y": 23 + }, + "id": 30, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_status_phase{cluster=~\"$cluster\",namespace=~\"$namespace\", phase=\"Running\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Pods Running", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "fixedColor": "#629e51", + "mode": "fixed" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 12, + "y": 185 + }, + "id": 33, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_status_phase{cluster=~\"$cluster\",namespace=~\"$namespace\", phase=\"Succeeded\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Pods Succeeded", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "fixedColor": "#629e51", + "mode": "fixed" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 18, + "y": 185 + }, + "id": 32, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_status_phase{cluster=~\"$cluster\",namespace=~\"$namespace\", phase=\"Failed\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Pods Failed", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "fixedColor": "#629e51", + "mode": "fixed" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 12, + "y": 188 + }, + "id": 31, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_status_phase{cluster=~\"$cluster\",namespace=~\"$namespace\", phase=\"Pending\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Pods Pending", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "fixedColor": "#629e51", + "mode": "fixed" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 18, + "y": 188 + }, + "id": 34, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_status_phase{cluster=~\"$cluster\",namespace=~\"$namespace\", phase=\"Unknown\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Pods Unknown", + "type": "stat" + } + ], + "title": "Pods", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 14 + }, + "id": 36, + "panels": [ + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 0, + "y": 250 + }, + "id": 38, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_container_status_running{namespace=~\"$namespace\",cluster=~\"$cluster\"})", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Containers Running", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 6, + "y": 250 + }, + "id": 39, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_container_status_waiting{namespace=~\"$namespace\",cluster=~\"$cluster\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Containers Waiting", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 12, + "y": 250 + }, + "id": 40, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_container_status_terminated{namespace=~\"$namespace\",cluster=~\"$cluster\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Containers Terminated", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 6, + "x": 18, + "y": 250 + }, + "id": 41, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(changes(kube_pod_container_status_restarts_total{namespace=~\"$namespace\",cluster=~\"$cluster\"}[30m]))", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Containers Restarts (Last 30 Minutes)", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 12, + "x": 0, + "y": 253 + }, + "id": 43, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_container_resource_requests_cpu_cores{namespace=~\"$namespace\",cluster=~\"$cluster\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "CPU Cores Requested by Containers", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 12, + "x": 12, + "y": 253 + }, + "id": 42, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_pod_container_resource_requests_memory_bytes{namespace=~\"$namespace\" ,cluster=~\"$cluster\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Memory Requested By Containers", + "type": "stat" + } + ], + "title": "Containers", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 15 + }, + "id": 80, + "panels": [ + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 24, + "x": 0, + "y": 25 + }, + "id": 82, + "options": { + "alertThreshold": true, + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "kube_hpa_status_current_replicas{cluster=~\"$cluster\",namespace=~\"$namespace\"}", + "instant": false, + "interval": "", + "legendFormat": "current_{{hpa}}", + "refId": "A" + }, + { + "expr": "kube_hpa_spec_max_replicas{cluster=~\"$cluster\",namespace=~\"$namespace\"}", + "instant": false, + "interval": "", + "legendFormat": "max_{{hpa}}", + "refId": "B" + }, + { + "expr": "kube_hpa_spec_min_replicas{cluster=~\"$cluster\",namespace=~\"$namespace\"}", + "instant": false, + "interval": "", + "legendFormat": "min_{{hpa}}", + "refId": "C" + } + ], + "title": "hpa", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 24, + "x": 0, + "y": 108 + }, + "id": 89, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_hpa_status_current_replicas{hpa=~\".*\"})", + "instant": false, + "interval": "", + "legendFormat": "kube_hpa_status_current_replicas", + "refId": "A" + } + ], + "title": "total-hpa-current", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "gridPos": { + "h": 10, + "w": 12, + "x": 0, + "y": 116 + }, + "id": 90, + "options": {}, + "pluginVersion": "7.1.1", + "targets": [ + { + "expr": "kube_hpa_status_current_replicas{hpa=~\".*\"} == kube_hpa_spec_max_replicas{hpa=~\".*\"}", + "instant": false, + "interval": "", + "legendFormat": "{{hpa}}", + "refId": "A" + } + ], + "title": "current==max", + "type": "timeseries" + }, + { + "fieldConfig": { + "defaults": { + "custom": {} + }, + "overrides": [] + }, + "gridPos": { + "h": 10, + "w": 12, + "x": 12, + "y": 116 + }, + "id": 92, + "options": {}, + "pluginVersion": "7.1.1", + "targets": [ + { + "expr": "kube_hpa_status_current_replicas{hpa=~\".*\"} == kube_hpa_spec_min_replicas{hpa=~\".*\"}", + "instant": false, + "interval": "", + "legendFormat": "{{hpa}}", + "refId": "B" + } + ], + "title": "current==min", + "type": "timeseries" + } + ], + "title": "HPA", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 16 + }, + "id": 45, + "panels": [ + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 8, + "x": 0, + "y": 26 + }, + "id": 47, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_job_status_succeeded{namespace=~\"$namespace\",cluster=~\"$cluster\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Jobs Succeeded", + "type": "stat" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "inspect": false + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 8, + "x": 8, + "y": 26 + }, + "id": 62, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "(time()-kube_job_status_completion_time{namespace=~\"$namespace\",cluster=~\"$cluster\"})/60/60/24>1", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{job_name}}", + "refId": "A" + } + ], + "title": "Job last success time from now", + "transparent": true, + "type": "table" + }, + { + "description": "", + "fieldConfig": { + "defaults": { + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "inspect": false + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 8, + "x": 16, + "y": 26 + }, + "id": 76, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "kube_job_status_failed{namespace=~\"$namespace\",cluster=~\"$cluster\"}>1", + "format": "time_series", + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{job_name}}", + "refId": "A" + } + ], + "title": "Job failed", + "transparent": true, + "type": "table" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 8, + "x": 0, + "y": 37 + }, + "id": 48, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_job_status_active{namespace=~\"$namespace\",cluster=~\"$cluster\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Jobs Active", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 3, + "w": 8, + "x": 0, + "y": 40 + }, + "id": 49, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_job_status_failed{namespace=~\"$namespace\",cluster=~\"$cluster\"})", + "format": "time_series", + "intervalFactor": 1, + "refId": "A" + } + ], + "title": "Jobs Failed", + "type": "stat" + } + ], + "title": "Jobs", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 17 + }, + "id": 64, + "panels": [ + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 6, + "x": 0, + "y": 27 + }, + "id": 66, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_persistentvolumeclaim_status_phase{phase=\"Bound\"}==1)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{persistentvolumeclaim=}}", + "refId": "A" + } + ], + "title": "pvc is bound", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 6, + "x": 6, + "y": 27 + }, + "id": 69, + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "sum(kube_persistentvolumeclaim_status_phase{phase!=\"Bound\"}==1)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{persistentvolumeclaim=}}", + "refId": "A" + } + ], + "title": "pvc is not bound", + "type": "stat" + }, + { + "fieldConfig": { + "defaults": { + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "inspect": false + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 27 + }, + "id": 67, + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "showHeader": true + }, + "pluginVersion": "12.0.0", + "targets": [ + { + "expr": "kube_persistentvolumeclaim_resource_requests_storage_bytes/1024/1024/1024", + "format": "time_series", + "instant": true, + "intervalFactor": 1, + "legendFormat": "{{persistentvolumeclaim}}", + "refId": "A" + } + ], + "title": "pvc", + "type": "table" + } + ], + "title": "PVC", + "type": "row" + } + ], + "preload": false, + "refresh": "", + "schemaVersion": 41, + "tags": ["kubernetes", "kubernetes-app"], + "templating": { + "list": [ + { + "current": { + "text": "", + "value": "" + }, + "definition": "label_values(kube_node_info, cluster)", + "includeAll": false, + "name": "cluster", + "options": [], + "query": "label_values(kube_node_info, cluster)", + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "current": { + "text": "All", + "value": "$__all" + }, + "definition": "label_values(kube_node_info{cluster=~\"$cluster\"}, node)", + "includeAll": true, + "name": "node", + "options": [], + "query": "label_values(kube_node_info{cluster=~\"$cluster\"}, node)", + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "current": { + "text": "All", + "value": "$__all" + }, + "definition": "", + "includeAll": true, + "name": "namespace", + "options": [], + "query": "label_values(kube_namespace_created{cluster=~\"$cluster\"}, namespace)", + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 2, + "name": "datasource", + "query": "Prometheus", + "skipUrlSync": true, + "type": "constant" + } + ] + }, + "time": { + "from": "now-24h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ] + }, + "timezone": "browser", + "title": "kube-state-metrics-v2", + "uid": "garysdevil-kube-state-metrics-v2", + "version": 1 +} diff --git a/rootfs/usr/share/grafana/dashboards/node.json b/rootfs/usr/share/grafana/dashboards/node.json new file mode 100644 index 0000000..f76826d --- /dev/null +++ b/rootfs/usr/share/grafana/dashboards/node.json @@ -0,0 +1,3613 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "target": { + "limit": 100, + "matchAny": false, + "tags": [], + "type": "dashboard" + }, + "type": "dashboard" + } + ] + }, + "description": "cpu memroy disk IO network temperature", + "editable": false, + "fiscalYearStartMonth": 0, + "gnetId": 8919, + "graphTooltip": 0, + "id": null, + "links": [ + { + "asDropdown": true, + "icon": "external link", + "includeVars": true, + "keepTime": true, + "tags": [], + "targetBlank": true, + "title": "More", + "tooltip": "", + "type": "dashboards", + "url": "" + } + ], + "liveNow": false, + "panels": [ + { + "collapsed": false, + "datasource": { + "type": "prometheus" + }, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 176, + "panels": [], + "title": "Basic info", + "type": "row" + }, + { + "datasource": { + "type": "prometheus" + }, + "description": "", + "editable": false, + "error": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 1 + }, + "id": 11, + "links": [], + "maxPerRow": 6, + "options": { + "code": { + "language": "plaintext", + "showLineNumbers": false, + "showMiniMap": false + }, + "content": "", + "mode": "html" + }, + "pluginVersion": "9.5.3", + "repeat": "node", + "repeatDirection": "h", + "style": {}, + "title": "$node", + "type": "text" + }, + { + "datasource": { + "type": "prometheus" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "decimals": 1, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(245, 54, 54, 0.9)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 1 + }, + { + "color": "rgba(50, 172, 45, 0.97)", + "value": 2 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 2, + "x": 0, + "y": 2 + }, + "hideTimeOverride": true, + "id": 15, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "text": {}, + "textMode": "auto" + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "expr": "time() - node_boot_time_seconds{instance=~\"$node\"}", + "format": "time_series", + "hide": false, + "instant": true, + "intervalFactor": 2, + "refId": "A", + "step": 40 + } + ], + "title": "uptime", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(245, 54, 54, 0.9)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 1 + }, + { + "color": "rgba(50, 172, 45, 0.97)", + "value": 2 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 2, + "x": 2, + "y": 2 + }, + "id": 14, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "text": {}, + "textMode": "auto" + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "expr": "count(count(node_cpu_seconds_total{instance=~\"$node\", mode='system'}) by (cpu))", + "format": "time_series", + "instant": true, + "intervalFactor": 1, + "legendFormat": "", + "refId": "A", + "step": 20 + } + ], + "title": "cpu cores", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(245, 54, 54, 0.9)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 2 + }, + { + "color": "rgba(50, 172, 45, 0.97)", + "value": 3 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 2, + "x": 4, + "y": 2 + }, + "id": 75, + "links": [], + "maxDataPoints": 100, + "options": { + "colorMode": "value", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "text": {}, + "textMode": "auto" + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "expr": "node_memory_MemTotal_bytes{instance=~\"$node\"}", + "format": "time_series", + "instant": true, + "intervalFactor": 1, + "legendFormat": "{{instance}}", + "refId": "A", + "step": 20 + } + ], + "title": "memory", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "decbytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 2, + "x": 6, + "y": 2 + }, + "hideTimeOverride": false, + "id": 171, + "links": [], + "maxDataPoints": 3, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "text": {}, + "textMode": "value" + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "exemplar": true, + "expr": "sum(node_filesystem_size_bytes {instance=~\"$node\",fstype=~\"ext4|xfs\"})", + "format": "table", + "hide": false, + "instant": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{mountpoint}}", + "refId": "A" + } + ], + "title": "system disk", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "decimals": 2, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 50 + }, + { + "color": "rgba(245, 54, 54, 0.9)", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 3, + "x": 8, + "y": 2 + }, + "id": 167, + "links": [], + "maxDataPoints": 100, + "options": { + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showThresholdLabels": false, + "showThresholdMarkers": true, + "text": {} + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "expr": "100 - (avg(irate(node_cpu_seconds_total{instance=~\"$node\",mode=\"idle\"}[5m])) * 100)", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A", + "step": 20 + } + ], + "title": "cpu usage（5m）", + "type": "gauge" + }, + { + "datasource": { + "type": "prometheus" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "decimals": 2, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 20 + }, + { + "color": "rgba(245, 54, 54, 0.9)", + "value": 50 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 3, + "x": 11, + "y": 2 + }, + "id": 20, + "links": [], + "maxDataPoints": 100, + "options": { + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showThresholdLabels": false, + "showThresholdMarkers": true, + "text": {} + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "expr": "avg(irate(node_cpu_seconds_total{instance=~\"$node\",mode=\"iowait\"}[5m])) * 100", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A", + "step": 20 + } + ], + "title": "CPU iowait（5m）", + "type": "gauge" + }, + { + "datasource": { + "type": "prometheus" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "decimals": 0, + "mappings": [], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 80 + }, + { + "color": "rgba(245, 54, 54, 0.9)", + "value": 90 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 3, + "x": 14, + "y": 2 + }, + "hideTimeOverride": false, + "id": 172, + "links": [], + "maxDataPoints": 100, + "options": { + "orientation": "auto", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showThresholdLabels": false, + "showThresholdMarkers": true, + "text": {} + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "expr": "(1 - (node_memory_MemAvailable_bytes{instance=~\"$node\"} / (node_memory_MemTotal_bytes{instance=~\"$node\"})))* 100", + "format": "time_series", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 20 + } + ], + "title": "Memory Usage", + "type": "gauge" + }, + { + "datasource": { + "type": "prometheus" + }, + "description": "Obtain the maximum partition through the variable maxmount.", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 70 + }, + { + "color": "rgba(245, 54, 54, 0.9)", + "value": 90 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 4, + "x": 17, + "y": 2 + }, + "id": 154, + "links": [], + "maxDataPoints": 100, + "options": { + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showThresholdLabels": false, + "showThresholdMarkers": true, + "text": {} + }, + "pluginVersion": "9.5.3", + "repeatDirection": "h", + "targets": [ + { + "expr": "100 - ((node_filesystem_avail_bytes{instance=~\"$node\",mountpoint=\"$maxmount\",fstype=~\"ext4|xfs\"} * 100) / node_filesystem_size_bytes {instance=~\"$node\",mountpoint=\"$maxmount\",fstype=~\"ext4|xfs\"})", + "format": "time_series", + "interval": "10s", + "intervalFactor": 1, + "refId": "A", + "step": 20 + } + ], + "title": "partion ($maxmount) usage", + "type": "gauge" + }, + { + "datasource": { + "type": "prometheus" + }, + "description": "", + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "decimals": 2, + "mappings": [], + "max": 10000, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 7000 + }, + { + "color": "rgba(245, 54, 54, 0.9)", + "value": 9000 + } + ] + }, + "unit": "short" + }, + "overrides": [] + }, + "gridPos": { + "h": 5, + "w": 3, + "x": 21, + "y": 2 + }, + "hideTimeOverride": false, + "id": 16, + "links": [], + "maxDataPoints": 100, + "options": { + "orientation": "horizontal", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showThresholdLabels": false, + "showThresholdMarkers": true, + "text": {} + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "expr": "node_filefd_allocated{instance=~\"$node\"}", + "format": "time_series", + "instant": false, + "interval": "10s", + "intervalFactor": 1, + "refId": "B" + } + ], + "title": "opend file descriptor", + "type": "gauge" + }, + { + "datasource": { + "type": "prometheus" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 2, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "15 minutes" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#6ED0E0", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "1 minute" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#BF1B00", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "5分钟" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#CCA300", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 7, + "w": 11, + "x": 0, + "y": 7 + }, + "id": 13, + "links": [], + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "8.2.3", + "targets": [ + { + "expr": "node_load1{instance=~\"$node\"}", + "format": "time_series", + "instant": false, + "interval": "10s", + "intervalFactor": 2, + "legendFormat": "{{instance}}_1m", + "metric": "", + "refId": "A", + "step": 20, + "target": "" + }, + { + "expr": "node_load5{instance=~\"$node\"}", + "format": "time_series", + "instant": false, + "interval": "10s", + "intervalFactor": 2, + "legendFormat": "{{instance}}_5m", + "refId": "B", + "step": 20 + }, + { + "expr": "node_load15{instance=~\"$node\"}", + "format": "time_series", + "instant": false, + "interval": "10s", + "intervalFactor": 2, + "legendFormat": "{{instance}}_15m", + "refId": "C", + "step": 20 + } + ], + "title": "Load average", + "type": "timeseries" + }, + { + "aliasColors": { + "Idle - Waiting for something to happen": "#052B51", + "guest": "#9AC48A", + "idle": "#052B51", + "iowait": "#EAB839", + "irq": "#BF1B00", + "nice": "#C15C17", + "sdb_I/O operations per second%": "#d683ce", + "softirq": "#E24D42", + "steal": "#FCE2DE", + "system": "#508642", + "user": "#5195CE", + "The proportion of disk spent on I/O operations": "#ba43a9" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "decimals": 2, + "description": "node_disk_io_time_seconds_total", + "fieldConfig": { + "defaults": { + "links": [] + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 13, + "x": 11, + "y": 7 + }, + "hiddenSeries": false, + "id": 7, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "maxPerRow": 6, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "avg(irate(node_cpu_seconds_total{instance=~\"$node\",mode=\"system\"}[2m])) by (instance)", + "format": "time_series", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{instance}}_System", + "refId": "A", + "step": 20 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "avg(irate(node_cpu_seconds_total{instance=~\"$node\",mode=\"user\"}[2m])) by (instance)", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "{{instance}}_User", + "range": true, + "refId": "B", + "step": 240 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "avg(irate(node_cpu_seconds_total{instance=~\"$node\",mode=\"idle\"}[2m])) by (instance)", + "format": "time_series", + "hide": true, + "intervalFactor": 2, + "legendFormat": "{{instance}}_Idle", + "range": true, + "refId": "F", + "step": 240 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "avg(irate(node_cpu_seconds_total{instance=~\"$node\",mode=\"iowait\"}[2m])) by (instance)", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "{{instance}}_Iowait", + "range": true, + "refId": "D", + "step": 240 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "irate(node_disk_io_time_seconds_total{instance=~\"$node\"}[2m])", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{instance}}_{{device}}_per second I/O%", + "range": true, + "refId": "C" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "CPU usage、disk I/O time（%）", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:513", + "format": "percentunit", + "label": "", + "logBase": 1, + "max": "1", + "show": true + }, + { + "$$hashKey": "object:514", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "fieldConfig": { + "defaults": { + "displayName": "", + "unit": "bytes" + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 11, + "x": 0, + "y": 14 + }, + "hiddenSeries": false, + "id": 207, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "exemplar": false, + "expr": "node_filesystem_size_bytes{instance=~'$node',fstype=~\"ext4|xfs\"}-node_filesystem_avail_bytes {instance=~'$node',fstype=~\"ext4|xfs\"}", + "format": "time_series", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{mountpoint}}_usage", + "refId": "C" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Partion usage", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "transformations": [], + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:3570", + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:3571", + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "datasource": { + "type": "prometheus" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "custom": { + "align": "auto", + "cellOptions": { + "type": "auto" + }, + "filterable": false, + "inspect": false + }, + "decimals": 2, + "displayName": "", + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": null + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "short" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "mountpoint" + }, + "properties": [ + { + "id": "displayName", + "value": "partion" + }, + { + "id": "unit", + "value": "bytes" + }, + { + "id": "decimals", + "value": 2 + }, + { + "id": "custom.align" + }, + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)" + } + ] + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Value #A" + }, + "properties": [ + { + "id": "displayName", + "value": "free" + }, + { + "id": "unit", + "value": "bytes" + }, + { + "id": "decimals", + "value": 2 + }, + { + "id": "custom.cellOptions", + "value": { + "type": "color-text" + } + }, + { + "id": "custom.align" + }, + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(245, 54, 54, 0.9)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 10000000000 + }, + { + "color": "rgba(50, 172, 45, 0.97)", + "value": 20000000000 + } + ] + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Value #B" + }, + "properties": [ + { + "id": "displayName", + "value": "usage" + }, + { + "id": "unit", + "value": "percentunit" + }, + { + "id": "decimals", + "value": 2 + }, + { + "id": "custom.cellOptions", + "value": { + "mode": "gradient", + "type": "color-background" + } + }, + { + "id": "custom.align" + }, + { + "id": "thresholds", + "value": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)", + "value": null + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 70 + }, + { + "color": "rgba(245, 54, 54, 0.9)", + "value": 90 + } + ] + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "Value #C" + }, + "properties": [ + { + "id": "displayName", + "value": "total" + }, + { + "id": "unit", + "value": "bytes" + }, + { + "id": "decimals", + "value": 1 + }, + { + "id": "custom.align" + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "fstype" + }, + "properties": [ + { + "id": "displayName", + "value": "file system" + }, + { + "id": "unit", + "value": "short" + }, + { + "id": "decimals", + "value": 2 + }, + { + "id": "custom.align" + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "instance" + }, + "properties": [ + { + "id": "displayName", + "value": "file system" + }, + { + "id": "decimals", + "value": 2 + }, + { + "id": "custom.align" + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "device" + }, + "properties": [ + { + "id": "decimals", + "value": 2 + }, + { + "id": "custom.align" + } + ] + } + ] + }, + "gridPos": { + "h": 7, + "w": 13, + "x": 11, + "y": 14 + }, + "id": 164, + "links": [], + "options": { + "cellHeight": "sm", + "footer": { + "countRows": false, + "fields": "", + "reducer": ["sum"], + "show": false + }, + "frameIndex": 0, + "showHeader": true, + "sortBy": [ + { + "desc": false, + "displayName": "file system" + } + ] + }, + "pluginVersion": "9.5.3", + "targets": [ + { + "exemplar": false, + "expr": "node_filesystem_size_bytes{instance=~'$node',fstype=~\"ext4|xfs\"}", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "C" + }, + { + "exemplar": true, + "expr": "node_filesystem_avail_bytes {instance=~'$node',fstype=~\"ext4|xfs\"}", + "format": "table", + "hide": true, + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + }, + { + "exemplar": false, + "expr": "1-(node_filesystem_free_bytes{instance=~'$node',fstype=~\"ext4|xfs\"} / node_filesystem_size_bytes{instance=~'$node',fstype=~\"ext4|xfs\"})", + "format": "table", + "hide": false, + "instant": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "", + "refId": "B" + } + ], + "title": "Disk usage", + "transformations": [ + { + "id": "merge", + "options": {} + }, + { + "id": "filterFieldsByName", + "options": { + "include": { + "names": [ + "fstype", + "mountpoint", + "Value #C", + "Value #A", + "Value #B", + "device" + ] + } + } + } + ], + "type": "table" + }, + { + "collapsed": true, + "datasource": { + "type": "prometheus" + }, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 21 + }, + "id": 211, + "panels": [ + { + "datasource": { + "type": "prometheus" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 2, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": true, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "links": [], + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [ + { + "matcher": { + "id": "byName", + "options": "MEM_Avaliable" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#6ED0E0", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "MEM_Cached" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#EF843C", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "MEM_Free" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#629E51", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "MEM_Total" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#6d1f62", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "MEM_Used" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#eab839", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "avaliable" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#9ac48a", + "mode": "fixed" + } + } + ] + }, + { + "matcher": { + "id": "byName", + "options": "totol memery" + }, + "properties": [ + { + "id": "color", + "value": { + "fixedColor": "#bf1b00", + "mode": "fixed" + } + } + ] + } + ] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 2 + }, + "id": 156, + "links": [], + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "8.2.3", + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "exemplar": true, + "expr": "node_memory_MemTotal_bytes{instance=~\"$node\"}", + "format": "time_series", + "hide": false, + "instant": false, + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{instance}}_TotalMem", + "refId": "A", + "step": 4 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "node_memory_MemTotal_bytes{instance=~\"$node\"} - node_memory_MemAvailable_bytes{instance=~\"$node\"}", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "{{instance}}_Used", + "range": true, + "refId": "B", + "step": 4 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "node_memory_MemFree_bytes{instance=~\"$node\"}", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "MEM_Free", + "range": true, + "refId": "C", + "step": 4 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "node_memory_Buffers_bytes{instance=~\"$node\"}", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "MEM_Buffers", + "range": true, + "refId": "D", + "step": 4 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "node_memory_Cached_bytes{instance=~\"$node\"}", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "MEM_Cached", + "range": true, + "refId": "E", + "step": 4 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "node_memory_MemAvailable_bytes{instance=~\"$node\"}", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{instance}}_Available", + "range": true, + "refId": "F", + "step": 4 + }, + { + "datasource": { + "type": "prometheus" + }, + "expr": "node_memory_MemTotal_bytes{instance=~\"$node\"} - (node_memory_Cached_bytes{instance=~\"$node\"} + node_memory_Buffers_bytes{instance=~\"$node\"} + node_memory_MemFree_bytes{instance=~\"$node\"})", + "format": "time_series", + "hide": true, + "intervalFactor": 1, + "refId": "G" + } + ], + "title": "Memory info", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "lineInterpolation": "linear", + "lineWidth": 2, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": true, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 2 + }, + "id": 206, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "8.2.3", + "targets": [ + { + "exemplar": true, + "expr": "node_memory_SwapFree_bytes{instance=~\"$node\"}", + "hide": false, + "interval": "", + "legendFormat": "Swap_Free", + "refId": "A" + }, + { + "exemplar": true, + "expr": "node_memory_SwapCached_bytes{instance=~\"$node\"}", + "hide": false, + "interval": "", + "legendFormat": "SwapCached", + "refId": "B" + }, + { + "exemplar": true, + "expr": "node_memory_SwapTotal_bytes{instance=~\"$node\"}", + "hide": false, + "interval": "", + "legendFormat": "SwapTotal", + "refId": "C" + } + ], + "title": "Swap memory", + "type": "timeseries" + } + ], + "title": "Memory", + "type": "row" + }, + { + "collapsed": true, + "datasource": { + "type": "prometheus" + }, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 22 + }, + "id": 209, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "", + "fieldConfig": { + "defaults": { + "links": [] + }, + "overrides": [] + }, + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 3 + }, + "height": "300", + "hiddenSeries": false, + "id": 157, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + { + "$$hashKey": "object:101", + "alias": "/.*_out transmit$/", + "transform": "negative-Y" + } + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "irate(node_network_receive_bytes_total{instance=~'$node',device!~'tap.*|veth.*|br.*|docker.*|virbr*|lo*|lxc.*|nodelocaldns|cilium.*'}[5m])*8", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}_{{device}}_in receive", + "range": true, + "refId": "A", + "step": 4 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "irate(node_network_transmit_bytes_total{instance=~'$node',device!~'tap.*|veth.*|br.*|docker.*|virbr*|lo*|lxc.*|nodelocaldns|cilium.*'}[5m])*8", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}_{{device}}_out transmit", + "range": true, + "refId": "B", + "step": 4 + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Network flow", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:85", + "format": "bps", + "label": "upload（-）/download（+）", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:86", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": { + "TCP": "#6ED0E0" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "", + "fieldConfig": { + "defaults": { + "links": [] + }, + "overrides": [] + }, + "fill": 0, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 3 + }, + "height": "300", + "hiddenSeries": false, + "id": 158, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "node_netstat_Tcp_CurrEstab{instance=~'$node'}", + "format": "time_series", + "hide": false, + "interval": "10s", + "intervalFactor": 1, + "legendFormat": "{{instance}}_ESTABLISHED", + "refId": "A", + "step": 20 + }, + { + "expr": "node_sockstat_TCP_tw{instance=~'$node'}", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{instance}}_TCP_tw", + "refId": "D" + }, + { + "expr": "irate(node_netstat_Tcp_ActiveOpens{instance=~'$node'}[1m])", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{instance}}_ActiveOpens", + "refId": "B" + }, + { + "expr": "irate(node_netstat_Tcp_PassiveOpens{instance=~'$node'}[1m])", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{instance}}_PassiveOpens", + "refId": "C" + }, + { + "expr": "node_sockstat_TCP_alloc{instance=~'$node'}", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{instance}}_TCP_alloc", + "refId": "E" + }, + { + "expr": "node_sockstat_TCP_inuse{instance=~'$node'}", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{instance}}_TCP_inuse", + "refId": "F" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "TCP connect info", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:269", + "format": "short", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:270", + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "title": "Network", + "type": "row" + }, + { + "collapsed": true, + "datasource": { + "type": "prometheus" + }, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 23 + }, + "id": 190, + "panels": [ + { + "aliasColors": { + "vda_write": "#6ED0E0" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "Reads completed: Reads completed per second for each disk partition\n\nWrites completed: \nIO now", + "fieldConfig": { + "defaults": { + "links": [] + }, + "overrides": [] + }, + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 24 + }, + "height": "300", + "hiddenSeries": false, + "id": 161, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + { + "$$hashKey": "object:386", + "alias": "/.*_read$/", + "transform": "negative-Y" + } + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "irate(node_disk_reads_completed_total{instance=~\"$node\"}[2m])", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{instance}}_{{device}}_read", + "range": true, + "refId": "A", + "step": 10 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "irate(node_disk_writes_completed_total{instance=~\"$node\"}[2m])", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "{{instance}}_{{device}}_write", + "range": true, + "refId": "B", + "step": 10 + }, + { + "datasource": { + "type": "prometheus" + }, + "expr": "node_disk_io_now{instance=~\"$node\"}", + "format": "time_series", + "hide": true, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{device}}", + "refId": "C" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "disk read and write（IOPS）", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:393", + "format": "iops", + "label": "read（-）/write（+）I/O ops/sec", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:394", + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": { + "vda": "#6ED0E0" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "Read time ms The number of seconds taken for each disk partition read operation\n\nWrite time ms\nIO time ms \nIO time weighted", + "fieldConfig": { + "defaults": { + "links": [] + }, + "overrides": [] + }, + "fill": 3, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 24 + }, + "height": "300", + "hiddenSeries": false, + "id": 160, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + { + "$$hashKey": "object:459", + "alias": "/,*_read$/", + "transform": "negative-Y" + } + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "irate(node_disk_io_time_seconds_total{instance=~\"$node\"}[2m])", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{device}}", + "range": true, + "refId": "A", + "step": 10 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "irate(node_disk_io_time_weighted_seconds_total{instance=~\"$node\"}[2m])", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{device}}_weighting", + "range": true, + "refId": "D" + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "irate(node_disk_read_time_seconds_total{instance=~\"$node\"}[2m])", + "format": "time_series", + "hide": false, + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{instance}}_{{device}}_read", + "range": true, + "refId": "B" + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "irate(node_disk_write_time_seconds_total{instance=~\"$node\"}[2m])", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{instance}}_{{device}}_write", + "range": true, + "refId": "C" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "disk io read write time", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:466", + "format": "s", + "label": "read（-）/write（+）", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:467", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "The number of read operations per second=[Δrd_ios/Δt]", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 31 + }, + "hiddenSeries": false, + "id": 194, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideZero": false, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "rate(node_disk_reads_completed_total{instance=~'$node'}[2m])", + "interval": "", + "legendFormat": "{{device}}", + "range": true, + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "r/s", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:1217", + "format": "short", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:1218", + "format": "bytes", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "The number of write operations per second=[Δwr_ios/Δt]", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 31 + }, + "hiddenSeries": false, + "id": 196, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "(rate(node_disk_writes_completed_total{instance=~\"$node\"}[2m]))", + "interval": "", + "legendFormat": "{{device}}", + "range": true, + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "w/s", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:1363", + "format": "short", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:1364", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "The average time required for each read operation=[ Δ Rd_ticks/ Δ Rd_ios not only includes the time required for hard disk device read operations, but also the waiting time in the kernel queue.", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 39 + }, + "hiddenSeries": false, + "id": 202, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "(rate(node_disk_read_time_seconds_total {instance=~'$node'}[2m])) /(rate(node_disk_reads_completed_total {instance=~'$node'}[2m]))", + "interval": "", + "legendFormat": "{{device}}", + "range": true, + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "r_await", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:2148", + "format": "s", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:2149", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "Average time required for each write operation=[ Δ Wr_ticks/ Δ Wr_ios not only includes the time required for hard disk device write operations, but also the waiting time in the kernel queue.", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 39 + }, + "hiddenSeries": false, + "id": 204, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "(rate(node_disk_write_time_seconds_total {instance=~'$node'}[2m])) /(rate(node_disk_writes_completed_total {instance=~'$node'}[2m]))", + "interval": "", + "legendFormat": "{{device}}", + "range": true, + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "w_await", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:2477", + "format": "s", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:2478", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 47 + }, + "hiddenSeries": false, + "id": 198, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "(rate(node_disk_read_bytes_total {instance=~'$node'}[2m]))", + "interval": "", + "legendFormat": "{{device}}", + "range": true, + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "rMiB/s", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:1622", + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:1623", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 47 + }, + "hiddenSeries": false, + "id": 200, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 2, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "(rate(node_disk_written_bytes_total {instance=~'$node'}[2m]))", + "interval": "", + "legendFormat": "{{device}}", + "range": true, + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "wKiB/s", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:1920", + "format": "bytes", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:1921", + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": { + "vda_write": "#6ED0E0" + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "", + "fieldConfig": { + "defaults": { + "links": [] + }, + "overrides": [] + }, + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 10, + "w": 24, + "x": 0, + "y": 55 + }, + "height": "300", + "hiddenSeries": false, + "id": 168, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + { + "alias": "/.*_read$/", + "transform": "negative-Y" + } + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "irate(node_disk_read_bytes_total{instance=~\"$node\"}[2m])", + "format": "time_series", + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{instance}}_{{device}}_read", + "range": true, + "refId": "A", + "step": 10 + }, + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "irate(node_disk_written_bytes_total{instance=~\"$node\"}[2m])", + "format": "time_series", + "hide": false, + "intervalFactor": 2, + "legendFormat": "{{instance}}_{{device}}_write", + "range": true, + "refId": "B", + "step": 10 + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Disk read and write capacity size", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "Bps", + "label": "read（-）/write（+）", + "logBase": 1, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "description": "The busy ratio of this hard drive device=[ Δ Io_ticks/ Δ t] The time ratio indicating that the device has I/O (i.e. non idle), regardless of how much I/O there is, only considering whether there is", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 8, + "w": 24, + "x": 0, + "y": 65 + }, + "hiddenSeries": false, + "id": 192, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideZero": false, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "sort": "max", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 2, + "points": false, + "renderer": "flot", + "repeat": "idc", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "rate(node_disk_io_time_seconds_total{instance=~\"$node\"}[2m])", + "hide": false, + "instant": false, + "interval": "", + "legendFormat": "{{device}}", + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "%util", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:744", + "format": "short", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:745", + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "title": "Disk read write", + "type": "row" + }, + { + "collapsed": true, + "datasource": { + "type": "prometheus" + }, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 24 + }, + "id": 213, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": { + "type": "prometheus" + }, + "fieldConfig": { + "defaults": { + "links": [] + }, + "overrides": [] + }, + "fill": 0, + "fillGradient": 0, + "gridPos": { + "h": 10, + "w": 24, + "x": 0, + "y": 25 + }, + "hiddenSeries": false, + "id": 169, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null as zero", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "9.5.3", + "pointradius": 0.5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": { + "type": "prometheus" + }, + "editorMode": "code", + "expr": "node_hwmon_temp_celsius{instance=~'$node'}", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "{{chip}}_{{sensor}}", + "range": true, + "refId": "A" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "cpu temperature", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "yaxes": [ + { + "$$hashKey": "object:3976", + "format": "celsius", + "logBase": 1, + "show": true + }, + { + "$$hashKey": "object:3977", + "format": "short", + "logBase": 1, + "show": true + } + ], + "yaxis": { + "align": false + } + } + ], + "title": "Cpu temperature", + "type": "row" + } + ], + "refresh": "", + "schemaVersion": 38, + "style": "dark", + "tags": ["instance"], + "templating": { + "list": [ + { + "auto": true, + "auto_count": 10, + "auto_min": "10s", + "current": { + "selected": true, + "text": "5s", + "value": "5s" + }, + "hide": 0, + "label": "interval", + "name": "interval", + "options": [ + { + "selected": false, + "text": "auto", + "value": "$__auto_interval_interval" + }, + { + "selected": true, + "text": "5s", + "value": "5s" + }, + { + "selected": false, + "text": "10s", + "value": "10s" + }, + { + "selected": false, + "text": "30s", + "value": "30s" + }, + { + "selected": false, + "text": "1m", + "value": "1m" + }, + { + "selected": false, + "text": "10m", + "value": "10m" + }, + { + "selected": false, + "text": "30m", + "value": "30m" + }, + { + "selected": false, + "text": "1h", + "value": "1h" + }, + { + "selected": false, + "text": "6h", + "value": "6h" + }, + { + "selected": false, + "text": "12h", + "value": "12h" + }, + { + "selected": false, + "text": "1d", + "value": "1d" + } + ], + "query": "5s,10s,30s,1m,10m,30m,1h,6h,12h,1d", + "refresh": 2, + "skipUrlSync": false, + "type": "interval" + }, + { + "allFormat": "glob", + "allValue": "", + "current": { + "selected": false, + "text": [], + "value": [] + }, + "datasource": { + "type": "prometheus" + }, + "definition": "label_values(node_exporter_build_info{},instance)", + "hide": 0, + "includeAll": false, + "label": "host instance", + "multi": true, + "multiFormat": "regex values", + "name": "node", + "options": [], + "query": { + "query": "label_values(node_exporter_build_info{},instance)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "current": { + "selected": false, + "text": "/", + "value": "/" + }, + "datasource": { + "type": "prometheus" + }, + "definition": "", + "hide": 2, + "includeAll": false, + "label": "", + "multi": false, + "name": "maxmount", + "options": [], + "query": { + "query": "query_result(topk(1,sort_desc (max(node_filesystem_size_bytes{instance=~'$node',fstype=~\"ext4|xfs\"}) by (mountpoint))))", + "refId": "Prometheus-maxmount-Variable-Query" + }, + "refresh": 2, + "regex": "/.*\\\"(.*)\\\".*/", + "skipUrlSync": false, + "sort": 0, + "tagValuesQuery": "", + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "datasource": {}, + "filters": [], + "hide": 0, + "label": "", + "name": "Filters", + "skipUrlSync": false, + "type": "adhoc" + } + ] + }, + "time": { + "from": "now-5m", + "to": "now" + }, + "timepicker": { + "now": true, + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": ["5m", "15m", "1h", "6h", "12h", "24h", "2d", "7d", "30d"] + }, + "timezone": "browser", + "title": "Node Metrics", + "uid": "SQlD2vTZq", + "version": 1, + "weekStart": "" +} diff --git a/rootfs/usr/share/grafana/oauth2/__init__.py b/rootfs/usr/share/grafana/oauth2/__init__.py new file mode 100644 index 0000000..e69de29 diff --git a/rootfs/usr/share/grafana/oauth2/alerting/pod_cpu.json b/rootfs/usr/share/grafana/oauth2/alerting/pod_cpu.json new file mode 100644 index 0000000..d74402f --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/alerting/pod_cpu.json @@ -0,0 +1,80 @@ +{ + "uid": "pod_cpu", + "title": "pod cpu usage over 80%", + "condition": "C", + "data": [ + { + "refId": "A", + "relativeTimeRange": { + "from": 600, + "to": 0 + }, + "datasourceUid": "prometheus_on_drycc", + "model": { + "editorMode": "code", + "expr": "(sum(rate(container_cpu_usage_seconds_total{image!=\"\"}[2m])) by (pod,namespace)/sum(kube_pod_container_resource_limits{resource=\"cpu\"}) by (pod,namespace) * 100) \u003e 80", + "instant": true, + "intervalMs": 1000, + "legendFormat": "__auto", + "maxDataPoints": 43200, + "range": false, + "refId": "A" + } + }, + { + "refId": "C", + "relativeTimeRange": { + "from": 0, + "to": 0 + }, + "datasourceUid": "__expr__", + "model": { + "conditions": [ + { + "evaluator": { + "params": [ + 0 + ], + "type": "gt" + }, + "operator": { + "type": "and" + }, + "query": { + "params": [ + "C" + ] + }, + "reducer": { + "params": [], + "type": "last" + }, + "type": "query" + } + ], + "datasource": { + "type": "__expr__", + "uid": "__expr__" + }, + "expression": "A", + "intervalMs": 1000, + "maxDataPoints": 43200, + "refId": "C", + "type": "threshold" + } + } + ], + "noDataState": "OK", + "execErrState": "KeepLast", + "for": "2m", + "ruleGroup": "middle", + "annotations": { + "description": "namespace: {{ $labels.namespace }}, pod: {{ $labels.pod }} 's cpu usage over {{$value }}%.", + "summary": "CPU Usage of Pod limit on has exceeded 80%." + }, + "isPaused": false, + "notification_settings": { + "receiver": "controller-alerts" + }, + "folderUID": "drycc" +} diff --git a/rootfs/usr/share/grafana/oauth2/alerting/pod_memory.json b/rootfs/usr/share/grafana/oauth2/alerting/pod_memory.json new file mode 100644 index 0000000..32c816d --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/alerting/pod_memory.json @@ -0,0 +1,81 @@ +{ + "uid": "pod_memory", + "title": "pod memory usage over 80%", + "condition": "C", + "data": [ + { + "refId": "A", + "relativeTimeRange": { + "from": 600, + "to": 0 + }, + "datasourceUid": "prometheus_on_drycc", + "model": { + "editorMode": "code", + "expr": "(sum(container_memory_working_set_bytes{pod!=\"\",container!=\"\"}) by (pod, namespace)) / (sum(container_spec_memory_limit_bytes{pod!=\"\"}) by (pod, namespace)) \u003e 0.8 and (sum(container_memory_working_set_bytes{pod!=\"\",container!=\"\"}) by (pod, namespace)) / (sum (container_spec_memory_limit_bytes{pod!=\"\"}) by (pod, namespace)) \u003c 2", + "instant": true, + "intervalMs": 1000, + "legendFormat": "__auto", + "maxDataPoints": 43200, + "range": false, + "refId": "A" + } + }, + { + "refId": "C", + "relativeTimeRange": { + "from": 0, + "to": 0 + }, + "datasourceUid": "__expr__", + "model": { + "conditions": [ + { + "evaluator": { + "params": [ + 0 + ], + "type": "gt" + }, + "operator": { + "type": "and" + }, + "query": { + "params": [ + "C" + ] + }, + "reducer": { + "params": [], + "type": "last" + }, + "type": "query" + } + ], + "datasource": { + "type": "__expr__", + "uid": "__expr__" + }, + "expression": "A", + "intervalMs": 1000, + "maxDataPoints": 43200, + "refId": "C", + "type": "threshold" + } + } + ], + "noDataState": "OK", + "execErrState": "KeepLast", + "for": "2m", + "ruleGroup": "middle", + "keepFiringFor": "2m", + "annotations": { + "description": "namespace: {{ $labels.namespace }}, pod: {{ $labels.pod }} 's memory usage over {{ $value }}%.", + "summary": "Memory Usage of Pod limit on has exceeded 80%." + }, + "isPaused": false, + "notification_settings": { + "receiver": "controller-alerts" + }, + "folderUID": "drycc" +} diff --git a/rootfs/usr/share/grafana/oauth2/alerting/pod_restart.json b/rootfs/usr/share/grafana/oauth2/alerting/pod_restart.json new file mode 100644 index 0000000..ef1c602 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/alerting/pod_restart.json @@ -0,0 +1,86 @@ +{ + "uid": "pod_restart", + "title": "pod restart", + "annotations": { + "description": "namespace: {{ $labels.namespace }}, pod: {{ $labels.pod }} restart.", + "summary": "Pod restart alert" + }, + "condition": "C", + "data": [ + { + "datasourceUid": "prometheus_on_drycc", + "model": { + "editorMode": "code", + "expr": "increase(kube_pod_container_status_restarts_total{pod!~\".*job.*\"}[30m]) > 0", + "instant": true, + "intervalMs": 1000, + "legendFormat": "__auto", + "maxDataPoints": 43200, + "range": false, + "refId": "A" + }, + "queryType": "", + "refId": "A", + "relativeTimeRange": { + "from": 600, + "to": 0 + } + }, + { + "datasourceUid": "__expr__", + "model": { + "conditions": [ + { + "evaluator": { + "params": [ + 0 + ], + "type": "gt" + }, + "operator": { + "type": "and" + }, + "query": { + "params": [ + "C" + ] + }, + "reducer": { + "params": [], + "type": "last" + }, + "type": "query" + } + ], + "datasource": { + "type": "__expr__", + "uid": "__expr__" + }, + "expression": "A", + "intervalMs": 1000, + "maxDataPoints": 43200, + "refId": "C", + "type": "threshold" + }, + "queryType": "", + "refId": "C", + "relativeTimeRange": { + "from": 0, + "to": 0 + } + } + ], + "execErrState": "OK", + "folderUID": "drycc", + "id": 1, + "isPaused": false, + "keep_firing_for": "2m", + "noDataState": "OK", + "notification_settings": { + "receiver": "controller-alerts" + }, + "orgID": 1, + "record": null, + "ruleGroup": "middle", + "for": "2m" +} diff --git a/rootfs/usr/share/grafana/oauth2/alerting/pod_start.json b/rootfs/usr/share/grafana/oauth2/alerting/pod_start.json new file mode 100644 index 0000000..119385c --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/alerting/pod_start.json @@ -0,0 +1,80 @@ +{ + "uid": "pod_start", + "title": "pod start fail", + "condition": "C", + "data": [ + { + "refId": "A", + "relativeTimeRange": { + "from": 600, + "to": 0 + }, + "datasourceUid": "prometheus_on_drycc", + "model": { + "editorMode": "code", + "expr": "kube_pod_status_phase{pod!~\".*job.*\", phase=~\"Failed|Unknown\"} == 1", + "instant": true, + "intervalMs": 1000, + "legendFormat": "__auto", + "maxDataPoints": 43200, + "range": false, + "refId": "A" + } + }, + { + "refId": "C", + "relativeTimeRange": { + "from": 0, + "to": 0 + }, + "datasourceUid": "__expr__", + "model": { + "conditions": [ + { + "evaluator": { + "params": [ + 0 + ], + "type": "gt" + }, + "operator": { + "type": "and" + }, + "query": { + "params": [ + "C" + ] + }, + "reducer": { + "params": [], + "type": "last" + }, + "type": "query" + } + ], + "datasource": { + "type": "__expr__", + "uid": "__expr__" + }, + "expression": "A", + "intervalMs": 1000, + "maxDataPoints": 43200, + "refId": "C", + "type": "threshold" + } + } + ], + "noDataState": "OK", + "execErrState": "KeepLast", + "for": "2m", + "ruleGroup": "middle", + "annotations": { + "description": "namespace: {{ $labels.namespace }}, pod: {{ $labels.pod }} start fail.", + "summary": "Pod start fail" + }, + "isPaused": false, + "notification_settings": { + "receiver": "controller-alerts" + }, + "folderUID": "drycc" +} diff --git a/rootfs/usr/share/grafana/oauth2/alerting/pvc_usage.json b/rootfs/usr/share/grafana/oauth2/alerting/pvc_usage.json new file mode 100644 index 0000000..822b176 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/alerting/pvc_usage.json @@ -0,0 +1,80 @@ +{ + "uid": "pvc_usage", + "title": "pvc usage over 80%", + "condition": "C", + "data": [ + { + "refId": "A", + "relativeTimeRange": { + "from": 600, + "to": 0 + }, + "datasourceUid": "prometheus_on_drycc", + "model": { + "editorMode": "code", + "expr": "kubelet_volume_stats_used_bytes{} / kubelet_volume_stats_capacity_bytes{} \u003e 0.8", + "instant": true, + "intervalMs": 1000, + "legendFormat": "__auto", + "maxDataPoints": 43200, + "range": false, + "refId": "A" + } + }, + { + "refId": "C", + "relativeTimeRange": { + "from": 0, + "to": 0 + }, + "datasourceUid": "__expr__", + "model": { + "conditions": [ + { + "evaluator": { + "params": [ + 0 + ], + "type": "gt" + }, + "operator": { + "type": "and" + }, + "query": { + "params": [ + "C" + ] + }, + "reducer": { + "params": [], + "type": "last" + }, + "type": "query" + } + ], + "datasource": { + "type": "__expr__", + "uid": "__expr__" + }, + "expression": "A", + "intervalMs": 1000, + "maxDataPoints": 43200, + "refId": "C", + "type": "threshold" + } + } + ], + "noDataState": "OK", + "execErrState": "KeepLast", + "for": "2m", + "annotations": { + "description": "pvc capacity {{$labels.namespace}} {{$labels.persistentvolumeclaim}} on {{$labels.topology_topolvm_io_node}} usage over {{$value }}%.", + "summary": "Pvc usage over 80%" + }, + "isPaused": false, + "notification_settings": { + "receiver": "controller-alerts" + }, + "folderUID": "drycc", + "ruleGroup": "middle" +} diff --git a/rootfs/usr/share/grafana/oauth2/credentials.py b/rootfs/usr/share/grafana/oauth2/credentials.py new file mode 100644 index 0000000..2af6af1 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/credentials.py @@ -0,0 +1,167 @@ +import sys +import logging +import argparse +import asyncio + +import httpx +import valkey +import valkey.asyncio +import valkey.exceptions + +from settings import settings + +logging.basicConfig( + level=logging.INFO, + format="%(asctime)s [%(levelname)s] %(message)s", +) +logger = logging.getLogger(__name__) + +# ── Constants ───────────────────────────────────────────────────── + +TOKEN_KEY = "datasource:oauth2:token" +INIT_LOCK_KEY = "datasource:oauth2:init_lock" +INIT_LOCK_TTL = 30 # Lock TTL 30 seconds +REFRESH_THRESHOLD = 7 * 24 * 3600 # Refresh when less than 7 days remain (for CronJob) + +# ── Function 1: used by main.py (get token, create if missing) ───────── + + +async def _fetch_and_save_token(valkey_client: valkey.asyncio.Valkey) -> str: + """ + Fetch a new token from Passport and save it to Valkey with TTL = expires_in. + Returns the access_token string. + """ + logger.info("Fetching new token from %s", settings.passport_token_url) + + async with httpx.AsyncClient() as client: + resp = await client.post( + settings.passport_token_url, + data={ + "grant_type": "client_credentials", + "client_id": settings.drycc_passport_key, + "client_secret": settings.drycc_passport_secret, + "scope": settings.drycc_passport_scopes, + }, + timeout=30.0, + ) + resp.raise_for_status() + token_response = resp.json() + + access_token = token_response["access_token"] + expires_in = token_response.get("expires_in", settings.session_ttl_fallback_seconds) + + await valkey_client.set(TOKEN_KEY, access_token, ex=expires_in) + + logger.info("Token saved (expires_in: %d seconds)", expires_in) + return access_token + + +async def introspect_token(token: str) -> bool: + """ + Check if the token is valid and has the required scopes via introspection. + """ + if not settings.drycc_passport_scopes: + return True + + introspect_url = settings.passport_token_url.replace("token/", "introspect/") + try: + async with httpx.AsyncClient() as client: + resp = await client.post( + introspect_url, + auth=(settings.drycc_passport_key, settings.drycc_passport_secret), + data={"token": token}, + timeout=5.0, + ) + if resp.status_code == 200: + data = resp.json() + if data.get("active"): + token_scopes = set(data.get("scope", "").split()) + required_scopes = set(settings.drycc_passport_scopes.split()) + return required_scopes == token_scopes + except Exception as e: + logger.info(f"Error introspecting token: {e}") + return False + + +async def get_token() -> str: + """ + Get the current access token. + + Flow: + 1. Fast path: read Valkey directly and return if valid + 2. Try to acquire a Valkey lock with blocking wait + 3. Double-check token and refresh if necessary + + Returns: access_token string + Raises: RuntimeError (if token cannot be obtained) + """ + valkey_client = await settings.get_valkey_client() + token = await valkey_client.get(TOKEN_KEY) + if token and await introspect_token(token): + return token + try: + async with valkey_client.lock(INIT_LOCK_KEY, timeout=30.0, blocking_timeout=30.0): + token = await valkey_client.get(TOKEN_KEY) + if token and await introspect_token(token): + return token + return await _fetch_and_save_token(valkey_client) + except valkey.exceptions.LockError: + raise RuntimeError("Timeout waiting for token refresh lock.") + + +# ── Function 2: CronJob entry point ────────────────────────────────────── + +async def main(): + """ + Asynchronous CronJob entry point. + """ + parser = argparse.ArgumentParser() + parser.add_argument( + "--force", + action="store_true", + help="Force refresh regardless of current token validity", + ) + args = parser.parse_args() + + logger.info("Token refresher started (force=%s)", args.force) + + valkey_client = await settings.get_valkey_client() + + try: + try: + await valkey_client.ping() + logger.info("Connected to Valkey") + except Exception as e: + logger.error("Failed to connect to Valkey: %s", e) + sys.exit(1) + + should_refresh = args.force + + if not should_refresh: + remaining = await valkey_client.ttl(TOKEN_KEY) + # ttl returns -2 if key missing, -1 if no expiration set + if remaining < 0: + logger.warning("No valid token found, fetching new token") + should_refresh = True + else: + remaining_days = remaining / 86400 + logger.info("Current token has %.1f days remaining", remaining_days) + if remaining > REFRESH_THRESHOLD: + logger.info("Token still valid, skip refresh") + return + logger.warning("Token expires in %.1f days, refreshing...", remaining_days) + should_refresh = True + + if should_refresh: + try: + await _fetch_and_save_token(valkey_client) + logger.info("Token refresh completed successfully") + except Exception as e: + logger.error("Token refresh failed: %s", e) + sys.exit(1) + finally: + await valkey_client.aclose() + + +if __name__ == "__main__": + asyncio.run(main()) diff --git a/rootfs/usr/share/grafana/oauth2/dashboards/drycc_applicaion.json b/rootfs/usr/share/grafana/oauth2/dashboards/drycc_applicaion.json new file mode 100644 index 0000000..2457a88 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/dashboards/drycc_applicaion.json @@ -0,0 +1,1147 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "description": "General Drycc App metrics.", + "editable": false, + "fiscalYearStartMonth": 0, + "graphTooltip": 0, + "id": 19, + "links": [], + "panels": [ + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 35, + "panels": [], + "title": "Logs", + "type": "row" + }, + { + "datasource": { + "type": "quickwit-quickwit-datasource", + "uid": "application_logs" + }, + "description": "", + "fieldConfig": { + "defaults": {}, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 19, + "x": 0, + "y": 1 + }, + "id": 34, + "options": { + "dedupStrategy": "none", + "enableInfiniteScrolling": false, + "enableLogDetails": true, + "prettifyLogMessage": true, + "showCommonLabels": false, + "showLabels": false, + "showTime": true, + "sortOrder": "Descending", + "wrapLogMessage": true + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "alias": "", + "bucketAggs": [ + { + "field": "", + "id": "2", + "settings": { + "interval": "auto" + }, + "type": "date_histogram" + } + ], + "datasource": { + "type": "quickwit-quickwit-datasource", + "uid": "application_logs" + }, + "metrics": [ + { + "id": "1", + "settings": {}, + "type": "logs" + } + ], + "query": "kubernetes.namespace_name:${app:singlequote} kubernetes.container_name:drycc-controller", + "refId": "A", + "timeField": "" + } + ], + "title": "", + "type": "logs" + }, + { + "datasource": { + "type": "quickwit-quickwit-datasource", + "uid": "application_logs" + }, + "description": "[Query Language](https://quickwit.io/docs/reference/query-language)", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 5, + "x": 19, + "y": 1 + }, + "id": 36, + "interval": "1m", + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "alias": "", + "bucketAggs": [ + { + "field": "", + "id": "2", + "settings": { + "interval": "auto" + }, + "type": "date_histogram" + } + ], + "datasource": { + "type": "quickwit-quickwit-datasource", + "uid": "application_logs" + }, + "metrics": [ + { + "id": "1", + "type": "count" + } + ], + "query": "kubernetes.namespace_name:${app:singlequote} kubernetes.container_name:drycc-controller", + "refId": "A", + "timeField": "" + } + ], + "title": "Logs volume", + "type": "timeseries" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 9 + }, + "id": 8, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 2, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 19, + "x": 0, + "y": 10 + }, + "id": 15, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum(rate(container_cpu_usage_seconds_total{namespace=~\"$app\",container!=\"\"}[5m])) by (pod) \n/\n(\n sum(container_spec_cpu_quota{namespace=~\"$app\",container!=\"\"}) by (pod) / 10000000\n)", + "hide": false, + "interval": "", + "legendFormat": "{{pod}}", + "range": true, + "refId": "A" + } + ], + "title": "CPU Load 5m", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 5, + "x": 19, + "y": 10 + }, + "id": 22, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "exemplar": false, + "expr": "sum((container_spec_cpu_quota{namespace=~\"$app\",container!=\"\"}/1000)/100)", + "instant": true, + "legendFormat": "Total CPU Cores", + "range": false, + "refId": "A" + } + ], + "title": "Total CPU Cores", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 24, + "x": 0, + "y": 18 + }, + "id": 21, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "(increase(container_cpu_cfs_throttled_periods_total{namespace=~\"$app\",container!=\"\" }[5m]))/\n(increase(container_cpu_cfs_periods_total{namespace=~\"$app\",container!=\"\"}[5m]))", + "instant": false, + "legendFormat": "{{pod}}", + "range": true, + "refId": "A" + } + ], + "title": "CPUThrottling ", + "type": "timeseries" + } + ], + "title": "CPU", + "type": "row" + }, + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 10 + }, + "id": 11, + "panels": [], + "title": "Memory", + "type": "row" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 19, + "x": 0, + "y": 11 + }, + "id": 14, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true, + "sortBy": "Last *", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "(container_memory_working_set_bytes{namespace=~\"$app\",container!=\"\"}) \n", + "hide": false, + "interval": "", + "legendFormat": "{{pod}}", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "editorMode": "code", + "expr": "", + "hide": true, + "instant": false, + "legendFormat": "Total", + "range": true, + "refId": "B" + } + ], + "title": "Used Memory", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 5, + "x": 19, + "y": 11 + }, + "id": 33, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum(container_spec_memory_limit_bytes{namespace=~\"$app\",container!=\"\"})", + "legendFormat": "__auto", + "range": true, + "refId": "A" + } + ], + "title": "Total Memory", + "type": "stat" + }, + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 18 + }, + "id": 2, + "panels": [], + "title": "Networking", + "type": "row" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "Bps" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 19 + }, + "id": 4, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum(rate(container_network_receive_bytes_total{namespace=~\"$app\"}[5m])) ", + "hide": false, + "interval": "", + "legendFormat": "Total", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "editorMode": "code", + "expr": "sum(rate(container_network_receive_bytes_total{namespace=~\"$app\"}[5m])) by (pod)", + "hide": false, + "instant": false, + "legendFormat": "__auto", + "range": true, + "refId": "B" + } + ], + "title": "Received", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "Bps" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 19 + }, + "id": 6, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum(rate(container_network_transmit_bytes_total{namespace=~\"$app\"}[5m]))", + "interval": "", + "legendFormat": "Total", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "editorMode": "code", + "expr": "sum(rate(container_network_transmit_bytes_total{namespace=~\"$app\"}[5m])) by (pod)", + "hide": false, + "instant": false, + "legendFormat": "__auto", + "range": true, + "refId": "B" + } + ], + "title": "Transmitted", + "type": "timeseries" + }, + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 27 + }, + "id": 27, + "panels": [], + "title": "Volume", + "type": "row" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 28 + }, + "id": 28, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum(rate(container_fs_reads_bytes_total{namespace=~\"$app\",container!=\"\"}[1m]))\n", + "hide": false, + "instant": false, + "legendFormat": "Total", + "range": true, + "refId": "A" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "editorMode": "code", + "expr": "sum(rate(container_fs_reads_bytes_total{namespace=~\"$app\",container!=\"\"}[1m])) by(pod)", + "hide": false, + "instant": false, + "legendFormat": "{{pod}}", + "range": true, + "refId": "B" + } + ], + "title": "IOPS Read", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 28 + }, + "id": 30, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum(rate(container_fs_writes_bytes_total{namespace=~\"$app\",container!=\"\"}[1m]))\n", + "legendFormat": "Total", + "range": true, + "refId": "A" + }, + { + "editorMode": "code", + "expr": "sum(rate(container_fs_writes_bytes_total{namespace=~\"$app\",container!=\"\"}[1m])) by(pod)", + "hide": false, + "instant": false, + "legendFormat": "__auto", + "range": true, + "refId": "B" + } + ], + "title": "IOPS Write", + "type": "timeseries" + } + ], + "preload": false, + "refresh": "60s", + "schemaVersion": 41, + "tags": ["drycc"], + "templating": { + "list": [ + { + "current": { + "text": "demo1", + "value": "demo1" + }, + "definition": "label_values(namespace)", + "includeAll": false, + "label": "app", + "name": "app", + "options": [], + "query": { + "qryType": 1, + "query": "label_values(namespace)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "", + "type": "query" + } + ] + }, + "time": { + "from": "now-3h", + "to": "now" + }, + "timepicker": {}, + "timezone": "", + "title": "Drycc App", + "uid": "cerzlzfbcevapp", + "version": 9 +} diff --git a/rootfs/usr/share/grafana/oauth2/dashboards/drycc_application_logs.json b/rootfs/usr/share/grafana/oauth2/dashboards/drycc_application_logs.json new file mode 100644 index 0000000..c34b157 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/dashboards/drycc_application_logs.json @@ -0,0 +1,303 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": { + "type": "grafana", + "uid": "-- Grafana --" + }, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "target": { + "limit": 100, + "matchAny": false, + "tags": [], + "type": "dashboard" + }, + "type": "dashboard" + } + ] + }, + "description": "General Drycc Logs metrics.", + "editable": false, + "fiscalYearStartMonth": 0, + "graphTooltip": 1, + "id": 61, + "links": [], + "panels": [ + { + "datasource": { + "type": "quickwit-quickwit-datasource", + "uid": "application_logs" + }, + "description": "[Query Language](https://quickwit.io/docs/reference/query-language)", + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "axisSoftMin": 0, + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 2, + "interval": "1m", + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": false + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "alias": "", + "bucketAggs": [ + { + "field": "", + "id": "2", + "settings": { + "interval": "auto" + }, + "type": "date_histogram" + } + ], + "datasource": { + "type": "quickwit-quickwit-datasource", + "uid": "application_logs" + }, + "metrics": [ + { + "id": "1", + "type": "count" + } + ], + "query": "kubernetes.namespace_name:${app:singlequote} kubernetes.pod_name:${instance:singlequote} ${query:text} -kubernetes.container_name:drycc-controller", + "refId": "A", + "timeField": "" + } + ], + "title": "Logs volume", + "type": "timeseries" + }, + { + "datasource": { + "type": "quickwit-quickwit-datasource", + "uid": "application_logs" + }, + "description": "", + "fieldConfig": { + "defaults": {}, + "overrides": [] + }, + "gridPos": { + "h": 17, + "w": 24, + "x": 0, + "y": 4 + }, + "id": 1, + "options": { + "dedupStrategy": "none", + "enableInfiniteScrolling": false, + "enableLogDetails": true, + "prettifyLogMessage": true, + "showCommonLabels": false, + "showLabels": false, + "showTime": true, + "sortOrder": "Descending", + "wrapLogMessage": true + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "alias": "", + "bucketAggs": [ + { + "field": "", + "id": "2", + "settings": { + "interval": "auto" + }, + "type": "date_histogram" + } + ], + "datasource": { + "type": "quickwit-quickwit-datasource", + "uid": "application_logs" + }, + "metrics": [ + { + "id": "1", + "settings": {}, + "type": "logs" + } + ], + "query": "kubernetes.namespace_name:${app:singlequote} kubernetes.pod_name:${instance:singlequote} ${query:text} -kubernetes.container_name:drycc-controller", + "refId": "A", + "timeField": "" + } + ], + "title": "Application Logs", + "type": "logs" + } + ], + "preload": false, + "refresh": "", + "schemaVersion": 41, + "tags": ["drycc"], + "templating": { + "list": [ + { + "current": { + "text": "", + "value": "" + }, + "label": "Query", + "name": "query", + "options": [ + { + "selected": true, + "text": "", + "value": "" + } + ], + "query": "", + "type": "textbox" + }, + { + "allValue": "*", + "current": { + "text": "All", + "value": "$__all" + }, + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "definition": "label_values(namespace)", + "includeAll": true, + "label": "App", + "name": "app", + "options": [], + "query": { + "qryType": 1, + "query": "label_values(namespace)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "allValue": "*", + "current": { + "text": "All", + "value": "$__all" + }, + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "definition": "label_values({namespace=\"$app\"},pod)", + "includeAll": true, + "label": "Instance", + "name": "instance", + "options": [], + "query": { + "qryType": 1, + "query": "label_values({namespace=\"$app\"},pod)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "", + "sort": 5, + "type": "query" + }, + { + "allowCustomValue": true, + "baseFilters": [], + "datasource": { + "type": "datasource", + "uid": "grafana" + }, + "description": "", + "filters": [], + "hide": 2, + "name": "filter", + "type": "adhoc" + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": {}, + "timezone": "", + "title": "Drycc Logs", + "uid": "aepz8ewtzq4g0f", + "version": 20 +} diff --git a/rootfs/usr/share/grafana/oauth2/dashboards/drycc_gateway.json b/rootfs/usr/share/grafana/oauth2/dashboards/drycc_gateway.json new file mode 100644 index 0000000..78f2d48 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/dashboards/drycc_gateway.json @@ -0,0 +1,1350 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "description": "General Drycc Gateway metrics.", + "editable": true, + "fiscalYearStartMonth": 0, + "graphTooltip": 0, + "id": 20, + "links": [], + "panels": [ + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 106, + "panels": [], + "title": "General", + "type": "row" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "fixedColor": "rgb(31, 120, 193)", + "mode": "fixed" + }, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": 0 + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "ops" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 6, + "x": 0, + "y": 1 + }, + "id": 12, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.1.1", + "targets": [ + { + "editorMode": "code", + "expr": "round(sum(irate(istio_requests_total{destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])), 0.001)", + "format": "time_series", + "intervalFactor": 1, + "range": true, + "refId": "A", + "step": 4 + } + ], + "title": "Client Request Volume", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": 0 + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "s" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 6, + "x": 6, + "y": 1 + }, + "id": 87, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.1.1", + "targets": [ + { + "editorMode": "code", + "expr": "(histogram_quantile(0.50, sum(irate(istio_request_duration_milliseconds_bucket{reporter=~\"$qrep\",destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (le)) / 1000) or histogram_quantile(0.50, sum(irate(istio_request_duration_seconds_bucket{reporter=~\"$qrep\",destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (le))", + "format": "time_series", + "interval": "", + "intervalFactor": 1, + "legendFormat": "P50", + "range": true, + "refId": "A" + }, + { + "expr": "(histogram_quantile(0.90, sum(irate(istio_request_duration_milliseconds_bucket{reporter=~\"$qrep\",destination_service=~\"$service\"}[$__rate_interval])) by (le)) / 1000) or histogram_quantile(0.90, sum(irate(istio_request_duration_seconds_bucket{reporter=~\"$qrep\",destination_service=~\"$service\"}[$__rate_interval])) by (le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "P90", + "refId": "B" + }, + { + "expr": "(histogram_quantile(0.99, sum(irate(istio_request_duration_milliseconds_bucket{reporter=~\"$qrep\",destination_service=~\"$service\"}[$__rate_interval])) by (le)) / 1000) or histogram_quantile(0.99, sum(irate(istio_request_duration_seconds_bucket{reporter=~\"$qrep\",destination_service=~\"$service\"}[$__rate_interval])) by (le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "P99", + "refId": "C" + } + ], + "title": "Client Request Duration", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "fixedColor": "rgb(31, 120, 193)", + "mode": "fixed" + }, + "mappings": [], + "noValue": "0", + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)", + "value": 0 + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 95 + }, + { + "color": "rgba(245, 54, 54, 0.9)", + "value": 99 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 6, + "x": 12, + "y": 1 + }, + "id": 98, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.1.1", + "targets": [ + { + "editorMode": "code", + "expr": "sum(irate(istio_requests_total{reporter=\"destination\",destination_service=~\"$service\",response_code!~\"5.*\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) / (sum(irate(istio_requests_total{reporter=\"destination\",destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) or on () vector(1))", + "format": "time_series", + "intervalFactor": 1, + "range": true, + "refId": "A" + } + ], + "title": "Server Success Rate (non-5xx responses)", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "fixedColor": "rgb(31, 120, 193)", + "mode": "fixed" + }, + "mappings": [], + "noValue": "0", + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "dark-red", + "value": 0 + }, + { + "color": "dark-yellow", + "value": 0.95 + }, + { + "color": "dark-green", + "value": 0.99 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [] + }, + "gridPos": { + "h": 4, + "w": 6, + "x": 18, + "y": 1 + }, + "id": 14, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.1.1", + "targets": [ + { + "editorMode": "code", + "expr": "sum(irate(istio_requests_total{reporter=~\"$qrep\",destination_service=~\"$service\",response_code!~\"5.*\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) / (sum(irate(istio_requests_total{reporter=~\"$qrep\",destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) or on () vector(1))", + "format": "time_series", + "intervalFactor": 1, + "range": true, + "refId": "A" + } + ], + "title": "Client Success Rate (non-5xx responses)", + "type": "stat" + }, + { + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 5 + }, + "id": 104, + "panels": [], + "title": "Client Workloads", + "type": "row" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": 0 + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "ops" + }, + "overrides": [ + { + "matcher": { + "id": "byValue", + "options": { + "op": "gte", + "reducer": "allIsNull", + "value": 0 + } + }, + "properties": [ + { + "id": "custom.hideFrom", + "value": { + "legend": true, + "tooltip": true, + "viz": false + } + } + ] + } + ] + }, + "gridPos": { + "h": 6, + "w": 12, + "x": 0, + "y": 6 + }, + "id": 25, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.1.1", + "targets": [ + { + "editorMode": "code", + "expr": "round(sum(irate(istio_requests_total{connection_security_policy=\"mutual_tls\",destination_service=~\"$service\",reporter=~\"$qrep\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace,response_code), 0.001)", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{ source_workload }}.{{ source_workload_namespace }} : {{ response_code }} ", + "range": true, + "refId": "A", + "step": 2 + }, + { + "editorMode": "code", + "expr": "round(sum(irate(istio_requests_total{connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\", reporter=~\"$qrep\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, response_code), 0.001)", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{ source_workload }}.{{ source_workload_namespace }} : {{ response_code }}", + "range": true, + "refId": "B", + "step": 2 + } + ], + "title": "Incoming Requests By Source And Response Code", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "max": 1.01, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": 0 + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percentunit" + }, + "overrides": [ + { + "matcher": { + "id": "byValue", + "options": { + "op": "gte", + "reducer": "allIsNull", + "value": 0 + } + }, + "properties": [ + { + "id": "custom.hideFrom", + "value": { + "legend": true, + "tooltip": true, + "viz": false + } + } + ] + } + ] + }, + "gridPos": { + "h": 6, + "w": 12, + "x": 12, + "y": 6 + }, + "id": 26, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.1.1", + "targets": [ + { + "editorMode": "code", + "expr": "sum(irate(istio_requests_total{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",response_code!~\"5.*\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace) / sum(irate(istio_requests_total{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace)", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{ source_workload }}.{{ source_workload_namespace }} (🔐mTLS)", + "range": true, + "refId": "A", + "step": 2 + }, + { + "editorMode": "code", + "expr": "sum(irate(istio_requests_total{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",response_code!~\"5.*\",namespace=\"$app\",pod=\"$instance\" }[$__rate_interval])) by (source_workload, source_workload_namespace) / sum(irate(istio_requests_total{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace)", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{ source_workload }}.{{ source_workload_namespace }}", + "range": true, + "refId": "B", + "step": 2 + } + ], + "title": "Incoming Success Rate (non-5xx responses) By Source", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": 0 + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "s" + }, + "overrides": [ + { + "matcher": { + "id": "byValue", + "options": { + "op": "gte", + "reducer": "allIsNull", + "value": 0 + } + }, + "properties": [ + { + "id": "custom.hideFrom", + "value": { + "legend": true, + "tooltip": true, + "viz": false + } + } + ] + } + ] + }, + "gridPos": { + "h": 6, + "w": 8, + "x": 0, + "y": 12 + }, + "id": 27, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.1.1", + "targets": [ + { + "editorMode": "code", + "expr": "(histogram_quantile(0.50, sum(irate(istio_request_duration_milliseconds_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le)) / 1000) or histogram_quantile(0.50, sum(irate(istio_request_duration_seconds_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P50 (🔐mTLS)", + "range": true, + "refId": "A", + "step": 2 + }, + { + "editorMode": "code", + "expr": "(histogram_quantile(0.90, sum(irate(istio_request_duration_milliseconds_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le)) / 1000) or histogram_quantile(0.90, sum(irate(istio_request_duration_seconds_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P90 (🔐mTLS)", + "range": true, + "refId": "B", + "step": 2 + }, + { + "editorMode": "code", + "expr": "(histogram_quantile(0.95, sum(irate(istio_request_duration_milliseconds_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le)) / 1000) or histogram_quantile(0.95, sum(irate(istio_request_duration_seconds_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P95 (🔐mTLS)", + "range": true, + "refId": "C", + "step": 2 + }, + { + "editorMode": "code", + "expr": "(histogram_quantile(0.99, sum(irate(istio_request_duration_milliseconds_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le)) / 1000) or histogram_quantile(0.99, sum(irate(istio_request_duration_seconds_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P99 (🔐mTLS)", + "range": true, + "refId": "D", + "step": 2 + }, + { + "editorMode": "code", + "expr": "(histogram_quantile(0.50, sum(irate(istio_request_duration_milliseconds_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le)) / 1000) or histogram_quantile(0.50, sum(irate(istio_request_duration_seconds_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P50", + "range": true, + "refId": "E", + "step": 2 + }, + { + "editorMode": "code", + "expr": "(histogram_quantile(0.90, sum(irate(istio_request_duration_milliseconds_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le)) / 1000) or histogram_quantile(0.90, sum(irate(istio_request_duration_seconds_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P90", + "range": true, + "refId": "F", + "step": 2 + }, + { + "editorMode": "code", + "expr": "(histogram_quantile(0.95, sum(irate(istio_request_duration_milliseconds_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le)) / 1000) or histogram_quantile(0.95, sum(irate(istio_request_duration_seconds_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P95", + "range": true, + "refId": "G", + "step": 2 + }, + { + "editorMode": "code", + "expr": "(histogram_quantile(0.99, sum(irate(istio_request_duration_milliseconds_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le)) / 1000) or histogram_quantile(0.99, sum(irate(istio_request_duration_seconds_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P99", + "range": true, + "refId": "H", + "step": 2 + } + ], + "title": "Incoming Request Duration By Source", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": 0 + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "decbytes" + }, + "overrides": [ + { + "matcher": { + "id": "byValue", + "options": { + "op": "gte", + "reducer": "allIsNull", + "value": 0 + } + }, + "properties": [ + { + "id": "custom.hideFrom", + "value": { + "legend": true, + "tooltip": true, + "viz": false + } + } + ] + } + ] + }, + "gridPos": { + "h": 6, + "w": 8, + "x": 8, + "y": 12 + }, + "id": 28, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.1.1", + "targets": [ + { + "editorMode": "code", + "expr": "histogram_quantile(0.50, sum(irate(istio_request_bytes_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P50 (🔐mTLS)", + "range": true, + "refId": "A", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.90, sum(irate(istio_request_bytes_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P90 (🔐mTLS)", + "range": true, + "refId": "B", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.95, sum(irate(istio_request_bytes_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P95 (🔐mTLS)", + "range": true, + "refId": "C", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.99, sum(irate(istio_request_bytes_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P99 (🔐mTLS)", + "range": true, + "refId": "D", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.50, sum(irate(istio_request_bytes_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P50", + "range": true, + "refId": "E", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.90, sum(irate(istio_request_bytes_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P90", + "range": true, + "refId": "F", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.95, sum(irate(istio_request_bytes_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P95", + "range": true, + "refId": "G", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.99, sum(irate(istio_request_bytes_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P99", + "range": true, + "refId": "H", + "step": 2 + } + ], + "title": "Incoming Request Size By Source", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green", + "value": 0 + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "decbytes" + }, + "overrides": [ + { + "matcher": { + "id": "byValue", + "options": { + "op": "gte", + "reducer": "allIsNull", + "value": 0 + } + }, + "properties": [ + { + "id": "custom.hideFrom", + "value": { + "legend": true, + "tooltip": true, + "viz": false + } + } + ] + } + ] + }, + "gridPos": { + "h": 6, + "w": 8, + "x": 16, + "y": 12 + }, + "id": 68, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.1.1", + "targets": [ + { + "editorMode": "code", + "expr": "histogram_quantile(0.50, sum(irate(istio_response_bytes_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P50 (🔐mTLS)", + "range": true, + "refId": "A", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.90, sum(irate(istio_response_bytes_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P90 (🔐mTLS)", + "range": true, + "refId": "B", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.95, sum(irate(istio_response_bytes_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P95 (🔐mTLS)", + "range": true, + "refId": "C", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.99, sum(irate(istio_response_bytes_bucket{reporter=~\"$qrep\", connection_security_policy=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P99 (🔐mTLS)", + "range": true, + "refId": "D", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.50, sum(irate(istio_response_bytes_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P50", + "range": true, + "refId": "E", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.90, sum(irate(istio_response_bytes_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P90", + "range": true, + "refId": "F", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.95, sum(irate(istio_response_bytes_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P95", + "range": true, + "refId": "G", + "step": 2 + }, + { + "editorMode": "code", + "expr": "histogram_quantile(0.99, sum(irate(istio_response_bytes_bucket{reporter=~\"$qrep\", connection_security_policy!=\"mutual_tls\", destination_service=~\"$service\",namespace=\"$app\",pod=\"$instance\"}[$__rate_interval])) by (source_workload, source_workload_namespace, le))", + "format": "time_series", + "hide": false, + "intervalFactor": 1, + "legendFormat": "{{source_workload}}.{{source_workload_namespace}} P99", + "range": true, + "refId": "H", + "step": 2 + } + ], + "title": "Response Size By Source", + "type": "timeseries" + } + ], + "preload": false, + "refresh": "60s", + "schemaVersion": 41, + "tags": [ + "drycc" + ], + "templating": { + "list": [ + { + "current": { + "text": "demo3", + "value": "demo3" + }, + "definition": "label_values(istio_requests_total,namespace)", + "description": "namespace ", + "label": "app", + "name": "app", + "options": [], + "query": { + "qryType": 1, + "query": "label_values(istio_requests_total,namespace)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "allowCustomValue": false, + "current": { + "text": "demo3-istio-58647c8789-ztcql", + "value": "demo3-istio-58647c8789-ztcql" + }, + "definition": "label_values(istio_requests_total{namespace=\"$app\"},pod)", + "description": "pod ", + "label": "instance", + "name": "instance", + "options": [], + "query": { + "qryType": 1, + "query": "label_values(istio_requests_total{namespace=\"$app\"},pod)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "allowCustomValue": false, + "current": { + "text": "All", + "value": [ + "$__all" + ] + }, + "definition": "query_result(sum(istio_requests_total{namespace=\"$app\",pod=\"$instance\"}) by (destination_service) or sum(istio_tcp_sent_bytes_total{namespace=\"$app\",pod=\"$instance\"}) by (destination_service))", + "includeAll": true, + "label": "Service", + "multi": true, + "name": "service", + "options": [], + "query": { + "qryType": 3, + "query": "query_result(sum(istio_requests_total{namespace=\"$app\",pod=\"$instance\"}) by (destination_service) or sum(istio_tcp_sent_bytes_total{namespace=\"$app\",pod=\"$instance\"}) by (destination_service))", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "/.*destination_service=\"([^\"]*).*/", + "type": "query" + }, + { + "allowCustomValue": false, + "current": { + "text": "source", + "value": "source" + }, + "hide": 2, + "includeAll": false, + "label": "Reporter", + "name": "qrep", + "options": [ + { + "selected": true, + "text": "source", + "value": "source" + } + ], + "query": "source", + "type": "custom" + } + ] + }, + "time": { + "from": "now-6h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ] + }, + "timezone": "", + "title": "Drycc Gateway", + "uid": "bb2f5242-9429-4f60-b520-20b7c70a0a9a", + "version": 4 +} \ No newline at end of file diff --git a/rootfs/usr/share/grafana/oauth2/dashboards/drycc_instance.json b/rootfs/usr/share/grafana/oauth2/dashboards/drycc_instance.json new file mode 100644 index 0000000..c27bb01 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/dashboards/drycc_instance.json @@ -0,0 +1,1193 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "description": "General Drycc Instance metrics.", + "editable": false, + "fiscalYearStartMonth": 0, + "graphTooltip": 0, + "id": 69, + "links": [], + "panels": [ + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 8, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "decimals": 2, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 20, + "x": 0, + "y": 1 + }, + "id": 15, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum(rate(container_cpu_usage_seconds_total{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}[10m])) by (pod) \n/\n(\n sum(container_spec_cpu_quota{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}) by (pod) / 10000000\n)", + "hide": false, + "interval": "", + "legendFormat": "10 minute load average", + "range": true, + "refId": "A" + }, + { + "editorMode": "code", + "expr": "sum(rate(container_cpu_usage_seconds_total{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}[5m])) by (pod) \n/\n(\n sum(container_spec_cpu_quota{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}) by (pod) / 10000000\n)", + "hide": false, + "instant": false, + "legendFormat": "5 minute load average", + "range": true, + "refId": "B" + } + ], + "title": "CPU Utilization", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "thresholds" + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 4, + "x": 20, + "y": 1 + }, + "id": 22, + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "exemplar": false, + "expr": "(container_spec_cpu_quota{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}/1000)/100", + "instant": true, + "legendFormat": "Total CPU Cores", + "range": false, + "refId": "A" + } + ], + "title": "Total CPU Cores", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 24, + "x": 0, + "y": 26 + }, + "id": 21, + "options": { + "legend": { + "calcs": [], + "displayMode": "list", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "(increase(container_cpu_cfs_throttled_periods_total{namespace=~\"$app\",container!=\"\",pod=~\"$instance\" }[5m]))\n /\n(increase(container_cpu_cfs_periods_total{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}[5m]))", + "instant": false, + "legendFormat": "throtting", + "range": true, + "refId": "A" + } + ], + "title": "CPUThrottling ", + "type": "timeseries" + } + ], + "title": "CPU", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 1 + }, + "id": 11, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 143 + }, + "id": 9, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true, + "sortBy": "Name", + "sortDesc": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum(container_memory_rss{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}) by (pod)", + "hide": false, + "interval": "", + "legendFormat": "rss", + "range": true, + "refId": "A" + }, + { + "editorMode": "code", + "expr": "sum(container_memory_cache{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}) by (pod)", + "hide": false, + "instant": false, + "legendFormat": "cache", + "range": true, + "refId": "B" + }, + { + "editorMode": "code", + "expr": "container_spec_memory_limit_bytes{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}", + "hide": false, + "instant": false, + "legendFormat": "total", + "range": true, + "refId": "C" + } + ], + "title": "Used Memory\t", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 143 + }, + "id": 14, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "(container_memory_working_set_bytes{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}*100) /container_spec_memory_limit_bytes{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}\n", + "hide": false, + "interval": "", + "legendFormat": "Used Memory", + "range": true, + "refId": "A" + } + ], + "title": "Used Memory", + "type": "timeseries" + } + ], + "title": "Memory", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 2 + }, + "id": 2, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "Bps" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 77 + }, + "id": 4, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum(rate(container_network_receive_bytes_total{namespace=~\"$app\",pod=~\"$instance\"}[5m])) by (pod)", + "hide": false, + "interval": "", + "legendFormat": "received", + "range": true, + "refId": "A" + } + ], + "title": "Received", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "Bps" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 77 + }, + "id": 6, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "multi", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum(rate(container_network_transmit_bytes_total{namespace=~\"$app\",pod=~\"$instance\"}[5m])) by (pod)", + "interval": "", + "legendFormat": "Transmitted", + "range": true, + "refId": "A" + } + ], + "title": "Transmitted", + "type": "timeseries" + } + ], + "title": "Networking", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 3 + }, + "id": 23, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 280 + }, + "id": 24, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "container_processes{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}", + "instant": false, + "legendFormat": "Process Count", + "range": true, + "refId": "A" + }, + { + "editorMode": "code", + "expr": "container_threads{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}", + "hide": false, + "instant": false, + "legendFormat": "Thread Count", + "range": true, + "refId": "B" + } + ], + "title": "", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 280 + }, + "id": 26, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "container_file_descriptors{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}\n", + "instant": false, + "legendFormat": "File Descriptors\t", + "range": true, + "refId": "A" + }, + { + "editorMode": "code", + "expr": "container_sockets{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}", + "hide": false, + "instant": false, + "legendFormat": "Sockets", + "range": true, + "refId": "B" + } + ], + "title": "", + "type": "timeseries" + } + ], + "title": "Process", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 4 + }, + "id": 27, + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 5 + }, + "id": 28, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum(rate(container_fs_reads_bytes_total{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}[1m]))\n", + "hide": false, + "instant": false, + "legendFormat": "read", + "range": true, + "refId": "A" + } + ], + "title": "IOPS Read", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 0, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "auto", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "none" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + } + }, + "overrides": [] + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 5 + }, + "id": 30, + "options": { + "legend": { + "calcs": ["lastNotNull"], + "displayMode": "table", + "placement": "right", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "rate(container_fs_writes_bytes_total{namespace=~\"$app\",container!=\"\",pod=~\"$instance\"}[1m])\n", + "legendFormat": "write", + "range": true, + "refId": "A" + } + ], + "title": "IOPS Write", + "type": "timeseries" + } + ], + "title": "TmpDisk", + "type": "row" + } + ], + "preload": false, + "refresh": "", + "schemaVersion": 41, + "tags": ["drycc"], + "templating": { + "list": [ + { + "definition": "label_values(namespace)", + "includeAll": false, + "label": "app", + "name": "app", + "options": [], + "query": { + "qryType": 1, + "query": "label_values(namespace)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "allowCustomValue": false, + + "definition": "label_values({namespace=\"$app\"},pod)", + "includeAll": false, + "label": "instance", + "name": "instance", + "options": [], + "query": { + "qryType": 1, + "query": "label_values({namespace=\"$app\"},pod)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "", + "type": "query" + } + ] + }, + "time": { + "from": "now-5m", + "to": "now" + }, + "timepicker": {}, + "timezone": "", + "title": "Drycc Instances", + "uid": "cerzlzfbcevpcc", + "version": 18 +} diff --git a/rootfs/usr/share/grafana/oauth2/dashboards/drycc_volume.json b/rootfs/usr/share/grafana/oauth2/dashboards/drycc_volume.json new file mode 100644 index 0000000..e12b014 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/dashboards/drycc_volume.json @@ -0,0 +1,454 @@ +{ + "annotations": { + "list": [ + { + "builtIn": 1, + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "description": "General Drycc Volume metrics.", + "editable": false, + "fiscalYearStartMonth": 0, + "graphTooltip": 0, + "id": 74, + "links": [], + "panels": [ + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "bytes" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 18, + "x": 0, + "y": 0 + }, + "id": 2, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max", "min"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "(\n sum without(instance, node) (kubelet_volume_stats_capacity_bytes{namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n -\n sum without(instance, node) (kubelet_volume_stats_available_bytes{cluster=\"$cluster\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n)\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "Used Space", + "range": true, + "refId": "A" + }, + { + "editorMode": "code", + "expr": "sum without(instance, node) (kubelet_volume_stats_available_bytes{ namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "Free Space", + "range": true, + "refId": "B" + } + ], + "title": "Volume Space Usage", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "decimals": 2, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)" + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 80 + }, + { + "color": "rgba(245, 54, 54, 0.9)", + "value": 90 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 6, + "x": 18, + "y": 0 + }, + "id": 3, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "(\n kubelet_volume_stats_capacity_bytes{ namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n -\n kubelet_volume_stats_available_bytes{ namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n)\n/\nkubelet_volume_stats_capacity_bytes{ namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n* 100\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "range": true, + "refId": "A" + } + ], + "title": "Volume Space Usage", + "type": "stat" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "color": { + "mode": "palette-classic" + }, + "custom": { + "axisBorderShow": false, + "axisCenteredZero": false, + "axisColorMode": "text", + "axisLabel": "", + "axisPlacement": "auto", + "barAlignment": 0, + "barWidthFactor": 0.6, + "drawStyle": "line", + "fillOpacity": 10, + "gradientMode": "none", + "hideFrom": { + "legend": false, + "tooltip": false, + "viz": false + }, + "insertNulls": false, + "lineInterpolation": "linear", + "lineWidth": 1, + "pointSize": 5, + "scaleDistribution": { + "type": "linear" + }, + "showPoints": "never", + "spanNulls": false, + "stacking": { + "group": "A", + "mode": "normal" + }, + "thresholdsStyle": { + "mode": "off" + } + }, + "mappings": [], + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "green" + }, + { + "color": "red", + "value": 80 + } + ] + }, + "unit": "none" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 18, + "x": 0, + "y": 7 + }, + "id": 4, + "options": { + "legend": { + "calcs": ["mean", "lastNotNull", "max", "min"], + "displayMode": "table", + "placement": "bottom", + "showLegend": true + }, + "tooltip": { + "hideZeros": false, + "mode": "single", + "sort": "none" + } + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "sum without(instance, node) (kubelet_volume_stats_inodes_used{ namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "Used inodes", + "range": true, + "refId": "A" + }, + { + "editorMode": "code", + "expr": "(\n sum without(instance, node) (kubelet_volume_stats_inodes{ namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n -\n sum without(instance, node) (kubelet_volume_stats_inodes_used{ namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n)\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": " Free inodes", + "range": true, + "refId": "B" + } + ], + "title": "Volume inodes Usage", + "type": "timeseries" + }, + { + "datasource": { + "type": "prometheus", + "uid": "prometheus_on_drycc" + }, + "fieldConfig": { + "defaults": { + "decimals": 2, + "mappings": [ + { + "options": { + "match": "null", + "result": { + "text": "N/A" + } + }, + "type": "special" + } + ], + "max": 100, + "min": 0, + "thresholds": { + "mode": "absolute", + "steps": [ + { + "color": "rgba(50, 172, 45, 0.97)" + }, + { + "color": "rgba(237, 129, 40, 0.89)", + "value": 80 + }, + { + "color": "rgba(245, 54, 54, 0.9)", + "value": 90 + } + ] + }, + "unit": "percent" + }, + "overrides": [] + }, + "gridPos": { + "h": 7, + "w": 6, + "x": 18, + "y": 7 + }, + "id": 5, + "maxDataPoints": 100, + "options": { + "colorMode": "none", + "graphMode": "none", + "justifyMode": "auto", + "orientation": "horizontal", + "percentChangeColorMode": "standard", + "reduceOptions": { + "calcs": ["lastNotNull"], + "fields": "", + "values": false + }, + "showPercentChange": false, + "textMode": "auto", + "wideLayout": true + }, + "pluginVersion": "12.0.2", + "targets": [ + { + "editorMode": "code", + "expr": "kubelet_volume_stats_inodes_used{ namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n/\nkubelet_volume_stats_inodes{ namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n* 100\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "range": true, + "refId": "A" + } + ], + "title": "Volume inodes Usage", + "type": "stat" + } + ], + "preload": false, + "refresh": "", + "schemaVersion": 41, + "tags": ["drycc"], + "templating": { + "list": [ + { + "current": { + "text": "", + "value": "" + }, + + "definition": "", + "hide": 2, + "includeAll": false, + "label": "cluster", + "name": "cluster", + "options": [], + "query": "", + "refresh": 2, + "regex": "", + "type": "query" + }, + { + "definition": "label_values(kubelet_volume_stats_available_bytes,namespace)", + "label": "Namespace", + "name": "namespace", + "options": [], + "query": { + "qryType": 1, + "query": "label_values(kubelet_volume_stats_available_bytes,namespace)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 1, + "regex": "", + "type": "query" + }, + { + "definition": "label_values(kubelet_volume_stats_available_bytes{namespace=\"$namespace\"},persistentvolumeclaim)", + "description": "", + "includeAll": false, + "label": "PersistentVolumeClaim", + "name": "volume", + "options": [], + "query": { + "qryType": 1, + "query": "label_values(kubelet_volume_stats_available_bytes{namespace=\"$namespace\"},persistentvolumeclaim)", + "refId": "PrometheusVariableQueryEditor-VariableQuery" + }, + "refresh": 2, + "regex": "", + "type": "query" + } + ] + }, + "time": { + "from": "now-12h", + "to": "now" + }, + "timepicker": {}, + "timezone": "", + "title": "Drycc Volume", + "uid": "ab92a8e8041bd567a", + "version": 3 +} diff --git a/rootfs/usr/share/grafana/oauth2/datasources/prometheus.json b/rootfs/usr/share/grafana/oauth2/datasources/prometheus.json new file mode 100644 index 0000000..edb10fa --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/datasources/prometheus.json @@ -0,0 +1,13 @@ +{ + "name": "Prometheus on Drycc", + "type": "prometheus", + "uid": "prometheus_on_drycc", + "url": "http://localhost:4000/proxy/prometheus/${workspace}", + "access": "proxy", + "isDefault": true, + "basicAuth": false, + "jsonData": { + "httpMethod": "POST", + "timeInterval": "${time_interval}" + } +} diff --git a/rootfs/usr/share/grafana/oauth2/datasources/quickwit.json b/rootfs/usr/share/grafana/oauth2/datasources/quickwit.json new file mode 100644 index 0000000..682aa01 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/datasources/quickwit.json @@ -0,0 +1,12 @@ +{ + "name": "Application Logs", + "type": "quickwit-quickwit-datasource", + "uid": "application_logs", + "url": "http://localhost:4000/proxy/quickwit/${workspace}", + "access": "proxy", + "basicAuth": false, + "jsonData": { + "index": "logs-*", + "logMessageField": "log" + } +} diff --git a/rootfs/usr/share/grafana/oauth2/hook/__init__.py b/rootfs/usr/share/grafana/oauth2/hook/__init__.py new file mode 100644 index 0000000..7cbad99 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/hook/__init__.py @@ -0,0 +1,11 @@ +from .grafana import ( + has_changed, init_org, sync_user, sync_role, sync_default, sync_datasources, sync_dashboards, + sync_alerting +) + +state_changed = has_changed +startup_hooks = [init_org] +login_hooks = [ + sync_user, sync_role, sync_default, sync_datasources, sync_dashboards, sync_alerting +] +destroy_hooks = [] diff --git a/rootfs/usr/share/grafana/oauth2/hook/grafana.py b/rootfs/usr/share/grafana/oauth2/hook/grafana.py new file mode 100644 index 0000000..cd1108c --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/hook/grafana.py @@ -0,0 +1,469 @@ +import logging +import time +import json +import httpx +from pathlib import Path +from string import Template +from psycopg import AsyncConnection +from settings import settings + +logger = logging.getLogger(__name__) + +DEFAULT_HEADERS = {"Content-Type": "application/json"} +DRYCC_GRAFANA_DASHBOARD = Path(__file__).resolve().parent.parent + +# Drycc Workspace role to Grafana role mapping +DRYCC_WORKSPACE_ROLE_MAPPING = {"admin": "Editor", "member": "Editor", "viewer": "Viewer"} +# Grafana default main organization ID, renamed to "drycc" by init_org on startup +DRYCC_ORG_ID = 1 + + +# ── Public functions (exported via __init__.py) ────────────────────────── + + +async def init_org(org_id=DRYCC_ORG_ID, name="drycc"): + headers = {"Content-Type": "application/json"} + async with httpx.AsyncClient() as client: + resp = await client.get(_api_url(f"/api/orgs/{org_id}", is_admin=True), headers=headers) + org = resp.json() + if org["name"] == name: + return + await client.put( + _api_url(f"/api/orgs/{org_id}", is_admin=True), + headers=headers, + json={"name": name}, + ) + + +async def has_changed(token: dict, userinfo: dict) -> bool: + """Detect drift between the cached userinfo / Grafana org state and the + live state in passport and the controller. + + Returns True when: + * any tracked userinfo field differs from passport's response, or + * the user's controller workspace membership count differs from the + number of workspace orgs they currently belong to in Grafana, or + * any workspace's mapped Grafana role differs from the role the user + currently has on the matching Grafana org. + Returns False when everything matches. + """ + access_token = token.get("access_token") + passport_headers = {"Authorization": f"Bearer {access_token}"} + + async with httpx.AsyncClient() as client: + passport_resp = await client.get( + settings.passport_userinfo_url, headers=passport_headers) + passport_resp.raise_for_status() + passport_userinfo = passport_resp.json() + + for field in ("preferred_username", "email", "is_superuser", "is_staff"): + if userinfo.get(field) != passport_userinfo.get(field): + return True + + workspaces = await _get_workspaces(access_token) + live_roles: dict[str, str] = {} + for ws in workspaces: + ws_name = ws["name"] + members = await _get_workspace_members(ws_name, access_token) + user_member = next( + (m for m in members if m["user"] == userinfo["preferred_username"]), None) + if user_member is None: + continue + live_roles[ws_name] = DRYCC_WORKSPACE_ROLE_MAPPING.get( + user_member["role"], "Viewer") + + grafana_resp = await client.get( + _api_url("/api/user/orgs"), headers=_api_headers({}, userinfo)) + grafana_resp.raise_for_status() + grafana_roles = { + org["name"]: org["role"] + for org in grafana_resp.json() + if org["name"] != "drycc" + } + + return live_roles != grafana_roles + + +async def sync_user(context: dict, token: dict, userinfo: dict): + async with httpx.AsyncClient() as client: + resp = await client.get(_api_url("/api/user"), headers=_api_headers(context, userinfo)) + user = resp.json() + context["user_id"] = user["id"] + if userinfo["is_superuser"]: + await client.put( + _api_url(f"/api/admin/users/{context['user_id']}/permissions", is_admin=True), + headers=_api_headers(context, userinfo), + json={"isGrafanaAdmin": userinfo["is_superuser"]}, + ) + + +async def sync_role(context: dict, token: dict, userinfo: dict): + """Sync user's Grafana Org memberships based on their workspace memberships.""" + access_token = token.get("access_token") + context["access_token"] = access_token + + workspace_orgs = await _build_workspace_orgs(userinfo, access_token) + + async with httpx.AsyncClient() as client: + resp = await client.get( + _api_url("/api/user/orgs"), headers=_api_headers(context, userinfo)) + current_orgs = resp.json() + workspace_org_ids = {info["org_id"] for info in workspace_orgs.values()} + current_org_map = {org["orgId"]: org["name"] for org in current_orgs} + + await _sync_workspace_org_memberships( + client, context, userinfo, workspace_orgs, current_org_map) + await _cleanup_stale_orgs( + client, context, userinfo, current_orgs, workspace_org_ids) + await _sync_drycc_org(client, context, userinfo, current_org_map) + + context["workspace_orgs"] = workspace_orgs + + +async def sync_default(context: dict, token: dict, userinfo: dict): + """Create default folder and alert configuration for each workspace org. + + Each workspace org gets a webhook contact point pointing back to this + service's /alerts/webhook endpoint, which fans out to passport messages + for workspace members with alerts=True. + """ + workspace_orgs = context.get("workspace_orgs", {}) + + for ws_name, ws_info in workspace_orgs.items(): + org_id = ws_info["org_id"] + + alertmanager_config = _build_alertmanager_config(ws_name) + + ctx = {**context, "org_id": org_id} + async with httpx.AsyncClient() as client: + await client.post( + _api_url("/api/folders"), + headers=_api_headers(ctx, userinfo), + json={"uid": "drycc", "title": "drycc"}, + ) + + await _upsert_alert_configuration(org_id, alertmanager_config) + + +async def sync_alerting(context: dict, token: dict, userinfo: dict): + """Create or update alert rules for each workspace org (idempotent). + + Alert rules are always created regardless of the alerts field. + The alerts field only controls notification channels (handled in sync_default). + """ + workspace_orgs = context.get("workspace_orgs", {}) + alerting_path = Path(__file__).resolve().parent.parent / "alerting" + + for _, ws_info in workspace_orgs.items(): + org_id = ws_info["org_id"] + ctx = {**context, "org_id": org_id} + + async with httpx.AsyncClient() as client: + for filepath in alerting_path.glob("*.json"): + with filepath.open() as f: + rule = json.load(f) + # Use PUT for idempotent upsert (POST would create duplicates) + resp = await client.put( + _api_url(f"/api/v1/provisioning/alert-rules/{rule['uid']}"), + headers=_api_headers(ctx, userinfo), + json=rule, + ) + if resp.status_code == 404: + # Rule doesn't exist yet, create it + await client.post( + _api_url("/api/v1/provisioning/alert-rules"), + headers=_api_headers(ctx, userinfo), + json=rule, + ) + + +async def sync_datasources(context: dict, token: dict, userinfo: dict): + """Create datasources for each workspace org with workspace-specific URLs. + + Datasource creation requires Org Admin permission. Workspace users are + only Editor/Viewer, so we must use Grafana admin basic auth combined with + X-Grafana-Org-Id to write into the target workspace org. + """ + workspace_orgs = context.get("workspace_orgs", {}) + datasources_path = Path(__file__).resolve().parent.parent / "datasources" + + for ws_name, ws_info in workspace_orgs.items(): + org_id = ws_info["org_id"] + headers = { + "Content-Type": "application/json", + "X-Grafana-Org-Id": str(org_id), + } + + async with httpx.AsyncClient() as client: + for filepath in datasources_path.glob("*.json"): + with filepath.open() as f: + template = Template(f.read()) + datasource = json.loads(template.substitute( + controller_url=settings.controller_base_url, + time_interval=settings.drycc_grafana_refresh, + workspace=ws_name, + )) + resp = await client.get( + _api_url(f"/api/datasources/name/{datasource['name']}", is_admin=True), + headers=headers) + if resp.status_code == 200: + existing = resp.json() + datasource["id"] = existing["id"] + datasource["version"] = existing["version"] + resp = await client.put( + _api_url(f"/api/datasources/uid/{datasource['uid']}", is_admin=True), + headers=headers, json=datasource) + resp.raise_for_status() + elif resp.status_code == 404: + resp = await client.post( + _api_url("/api/datasources", is_admin=True), + headers=headers, json=datasource) + resp.raise_for_status() + else: + raise ValueError( + f"grafana returned an unexpected status: {resp.status_code}" + ) + + +async def sync_dashboards(context: dict, token: dict, userinfo: dict): + """Create dashboards for each workspace org.""" + workspace_orgs = context.get("workspace_orgs", {}) + dashboards_path = Path(__file__).resolve().parent.parent / "dashboards" + + for ws_name, ws_info in workspace_orgs.items(): + org_id = ws_info["org_id"] + ctx = {**context, "org_id": org_id} + + async with httpx.AsyncClient() as client: + for filepath in dashboards_path.glob("*.json"): + with filepath.open() as f: + dashboard = json.load(f) + dashboard.update({"id": None, "refresh": settings.drycc_grafana_refresh}) + await client.post( + _api_url("/api/dashboards/db"), + headers=_api_headers(ctx, userinfo), + json={ + "dashboard": dashboard, + "folderUid": "drycc", + "overwrite": True, + }, + ) + + +# ── Private functions ──────────────────────────────────────────────────── + + +def _api_url(url_path, is_admin=False): + if is_admin: + return "http://{}:{}@localhost:{}{}".format( + settings.gf_security_admin_user, + settings.gf_security_admin_password, + settings.gf_server_http_port, + url_path + ) + return "http://localhost:{}{}".format(settings.gf_server_http_port, url_path) + + +def _api_headers(context: dict, userinfo): + headers = {"Content-Type": "application/json"} + headers["Remote-User"] = userinfo["preferred_username"] + headers["Remote-Name"] = userinfo["preferred_username"] + headers["Remote-Email"] = userinfo["email"] + if "org_id" in context: + headers["X-Grafana-Org-Id"] = str(context["org_id"]) + return headers + + +def _get_drycc_role(userinfo: dict) -> str | None: + if userinfo["is_superuser"]: + return "Editor" + elif userinfo["is_staff"]: + return "Viewer" + return None + + +async def _get_workspaces(access_token: str) -> list: + """Call Controller API to get user's workspaces.""" + headers = {"Authorization": f"Bearer {access_token}"} + async with httpx.AsyncClient() as client: + resp = await client.get( + f"{settings.controller_base_url}/v2/workspaces", headers=headers) + resp.raise_for_status() + return resp.json().get("results", []) + + +async def _get_workspace_members(workspace_id: str, access_token: str) -> list: + """Call Controller API to get workspace members.""" + headers = {"Authorization": f"Bearer {access_token}"} + async with httpx.AsyncClient() as client: + resp = await client.get( + f"{settings.controller_base_url}/v2/workspaces/{workspace_id}/members", + headers=headers) + resp.raise_for_status() + return resp.json().get("results", []) + + +async def _get_or_create_org(name: str) -> int: + """Find or create a Grafana Org by name. Returns org_id.""" + headers = {"Content-Type": "application/json"} + async with httpx.AsyncClient() as client: + # Try to find existing org by name + resp = await client.get( + _api_url(f"/api/orgs/name/{name}", is_admin=True), headers=headers) + if resp.status_code == 200: + return resp.json()["id"] + # Create new org (handle race condition: another request may have created it) + resp = await client.post( + _api_url("/api/orgs", is_admin=True), headers=headers, + json={"name": name}) + if resp.status_code == 409: + # Org already exists (race condition), fetch it + resp = await client.get( + _api_url(f"/api/orgs/name/{name}", is_admin=True), headers=headers) + return resp.json()["id"] + resp.raise_for_status() + return resp.json()["orgId"] + + +async def _build_workspace_orgs(userinfo: dict, access_token: str) -> dict: + """Build workspace org info by fetching workspaces and their memberships. + + Returns: {workspace_id: {"org_id": int, "role": str}} + """ + workspace_orgs = {} + try: + workspaces = await _get_workspaces(access_token) + except httpx.HTTPError as e: + logger.warning("Failed to fetch workspaces for %s: %s", userinfo["preferred_username"], e) + return workspace_orgs + + for ws in workspaces: + workspace_id = ws["id"] + try: + members = await _get_workspace_members(workspace_id, access_token) + except httpx.HTTPError as e: + logger.warning("Failed to fetch members for workspace %s: %s", workspace_id, e) + continue + user_member = next( + (m for m in members if m["user"] == userinfo["preferred_username"]), None) + if user_member is None: + continue + + org_id = await _get_or_create_org(workspace_id) + workspace_orgs[workspace_id] = { + "org_id": org_id, + "role": user_member["role"], + } + return workspace_orgs + + +async def _sync_workspace_org_memberships( + client: httpx.AsyncClient, context: dict, userinfo: dict, + workspace_orgs: dict, current_org_map: dict, +): + """Add/update user's membership in each workspace org.""" + for ws_name, ws_info in workspace_orgs.items(): + org_id = ws_info["org_id"] + grafana_role = DRYCC_WORKSPACE_ROLE_MAPPING.get(ws_info["role"], "Viewer") + + if org_id in current_org_map: + await client.patch( + _api_url(f"/api/orgs/{org_id}/users/{context['user_id']}", is_admin=True), + headers=_api_headers(context, userinfo), + json={"role": grafana_role}, + ) + else: + await client.post( + _api_url(f"/api/orgs/{org_id}/users", is_admin=True), + headers=_api_headers(context, userinfo), + json={ + "loginOrEmail": userinfo["preferred_username"], + "role": grafana_role, + }, + ) + + +async def _cleanup_stale_orgs( + client: httpx.AsyncClient, context: dict, userinfo: dict, + current_orgs: list, workspace_org_ids: set, +): + """Remove user from orgs they no longer belong to (excluding drycc org).""" + for org in current_orgs: + org_id = org["orgId"] + if org_id == DRYCC_ORG_ID: + continue + if org_id not in workspace_org_ids: + await client.delete( + _api_url(f"/api/orgs/{org_id}/users/{context['user_id']}", is_admin=True), + headers=_api_headers(context, userinfo), + ) + + +async def _sync_drycc_org( + client: httpx.AsyncClient, context: dict, userinfo: dict, + current_org_map: dict, +): + """Handle drycc org membership for superusers/staff.""" + drycc_role = _get_drycc_role(userinfo) + has_drycc_org = DRYCC_ORG_ID in current_org_map + if drycc_role: + if has_drycc_org: + await client.patch( + _api_url(f"/api/orgs/{DRYCC_ORG_ID}/users/{context['user_id']}", is_admin=True), + headers=_api_headers(context, userinfo), + json={"role": drycc_role}, + ) + else: + await client.post( + _api_url(f"/api/orgs/{DRYCC_ORG_ID}/users", is_admin=True), + headers=_api_headers(context, userinfo), + json={"loginOrEmail": userinfo["preferred_username"], "role": drycc_role}, + ) + elif has_drycc_org: + await client.delete( + _api_url(f"/api/orgs/{DRYCC_ORG_ID}/users/{context['user_id']}", is_admin=True), + headers=_api_headers(context, userinfo), + ) + + +def _build_alertmanager_config(ws_name: str) -> str: + receivers = [{ + "name": "controller-alerts", + "grafana_managed_receiver_configs": [{ + "uid": "", + "name": "controller alerts webhook", + "type": "webhook", + "settings": { + "url": f"http://localhost:4000/alerts/webhook?workspace={ws_name}", + "httpMethod": "POST", + }, + }], + }] + return json.dumps({ + "alertmanager_config": { + "route": { + "receiver": "controller-alerts", + "group_by": ["grafana_folder", "alertname"] + }, + "receivers": receivers + } + }) + + +async def _upsert_alert_configuration(org_id: int, config: str): + """Insert or update alert configuration for an org using parameterized query.""" + async with await AsyncConnection.connect(settings.gf_database_url) as conn: + async with conn.cursor() as cursor: + await cursor.execute( + """ + INSERT INTO alert_configuration ( + alertmanager_configuration, configuration_version, created_at, "default", org_id + ) VALUES (%s, %s, %s, %s, %s) + ON CONFLICT (org_id) DO UPDATE + SET alertmanager_configuration = EXCLUDED.alertmanager_configuration, + configuration_version = EXCLUDED.configuration_version, + created_at = EXCLUDED.created_at + """, + (config, "v1", int(time.time()), True, org_id), + ) + await conn.commit() diff --git a/rootfs/usr/share/grafana/oauth2/main.py b/rootfs/usr/share/grafana/oauth2/main.py new file mode 100644 index 0000000..0726ba4 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/main.py @@ -0,0 +1,234 @@ +import json +import random +import string +import argparse +import uuid +import httpx +from contextlib import asynccontextmanager +from fastapi import FastAPI, Request, HTTPException +from fastapi.responses import JSONResponse, RedirectResponse, Response, StreamingResponse +from authlib.integrations.starlette_client import OAuth +from starlette.middleware.sessions import SessionMiddleware + +from hook import state_changed, startup_hooks, login_hooks, destroy_hooks +from settings import settings +from credentials import get_token + + +def randstr(k=32): + return ''.join(random.choices(string.ascii_letters + string.digits, k=k)) + + +parser = argparse.ArgumentParser() +parser.add_argument('--port', default="8000", help='specify alternate port') +parser.add_argument('--bind', default="0.0.0.0", help='specify alternate bind address') +parser.add_argument('--client-id', required=True, help='the OAuth Client ID') +parser.add_argument('--client-secret', required=True, help='the OAuth Client secret') +parser.add_argument('--oidc-issuer-url', required=True, help='OpenID Connect issuer URL') +parser.add_argument('--cookie-name', default="_oauth2_proxy", help='the name of the cookie') +parser.add_argument('--cookie-secret', default=randstr, help='the seed string for secure cookies') +args = parser.parse_args() + + +http_client = None + +SESSION_KEY_PREFIX = "oauth2:session:" +STATE_CHECK_KEY_PREFIX = "oauth2:state_check:" + + +@asynccontextmanager +async def lifespan(app: FastAPI): + global http_client + http_client = httpx.AsyncClient(timeout=30.0) + + try: + await get_token() + except Exception as e: + raise RuntimeError(f"Failed to initialize token: {e}") + + for startup_hook in startup_hooks: + await startup_hook() + yield + await http_client.aclose() + for destroy_hook in destroy_hooks: + await destroy_hook() + + +app = FastAPI(lifespan=lifespan) +oauth = OAuth() +oauth.register( + name='oidc', + client_id=args.client_id, + client_secret=args.client_secret, + client_kwargs={'scope': 'openid profile email'}, + server_metadata_url=args.oidc_issuer_url + "/.well-known/openid-configuration", +) +app.add_middleware( + SessionMiddleware, secret_key=args.cookie_secret, session_cookie=args.cookie_name +) + + +@app.get("/oauth2/healthz") +async def healthz(): + return {"status": "ok"} + + +@app.get("/oauth2/sign_in") +async def oauth2_sign_in(request: Request): + sid = request.session.get("sid") + redirect_to = request.query_params.get("redirect", "/") + if sid: + valkey_client = await settings.get_valkey_client() + try: + key = f"{SESSION_KEY_PREFIX}{sid}" + if await valkey_client.get(key): + return RedirectResponse(redirect_to) + finally: + await valkey_client.aclose() + redirect_url = request.url_for("oauth2_callback").replace_query_params(redirect=redirect_to) + return await oauth.oidc.authorize_redirect(request, redirect_url) + + +@app.get("/oauth2/callback") +async def oauth2_callback(request: Request): + token = await oauth.oidc.authorize_access_token(request) + if 'userinfo' not in token: + userinfo = await oauth.oidc.userinfo(token=token) + else: + token.pop("id_token", None) + userinfo = token.pop("userinfo") + + sid = uuid.uuid4().hex + valkey_client = await settings.get_valkey_client() + try: + key = f"{SESSION_KEY_PREFIX}{sid}" + payload = json.dumps({"token": token, "user": userinfo}) + expires_in = token.get("expires_in") or settings.session_ttl_fallback_seconds + await valkey_client.set(key, payload, ex=expires_in) + finally: + await valkey_client.aclose() + request.session['sid'] = sid + + context = {} + for login_hook in login_hooks: + await login_hook(context, token, userinfo) + return RedirectResponse(url='/') + + +@app.get("/oauth2/userinfo") +async def oauth2_userinfo(request: Request): + sid = request.session.get("sid") + if not sid: + return JSONResponse(status_code=401, content={"error": "Unauthorized"}) + + session_key = f"{SESSION_KEY_PREFIX}{sid}" + state_check_key = f"{STATE_CHECK_KEY_PREFIX}{sid}" + valkey_client = await settings.get_valkey_client() + try: + raw = await valkey_client.get(session_key) + if not raw: + request.session.pop("sid", None) + return JSONResponse(status_code=401, content={"error": "Unauthorized"}) + + payload = json.loads(raw) + token, userinfo = payload["token"], payload["user"] + + # SET NX EX guarantees only one request per cooldown window runs the + # expensive upstream check, even when Grafana fans out concurrent + # /oauth2/userinfo calls. + cooldown = settings.state_check_cooldown_seconds + check_due = cooldown <= 0 or bool( + await valkey_client.set(state_check_key, "1", ex=cooldown, nx=True) + ) + if check_due and await state_changed(token, userinfo): + await valkey_client.delete(session_key) + request.session.pop("sid", None) + return JSONResponse(status_code=401, content={"error": "Unauthorized"}) + finally: + await valkey_client.aclose() + + headers = { + "Remote-User": userinfo['preferred_username'], + "Remote-Name": userinfo['preferred_username'], + "Remote-Email": userinfo['email'], + } + return JSONResponse(content=userinfo, headers=headers) + + +# ── Proxy routes (zero-overhead version) ─────────────────────────── + +async def _proxy_request(request: Request, url: str): + """Minimal proxy: get the token directly from token.py and use it.""" + try: + token = await get_token() + except RuntimeError as e: + raise HTTPException(status_code=503, detail=str(e)) + + headers = dict(request.headers) + headers.pop("host", None) + headers["Authorization"] = f"Bearer {token}" + resp = await http_client.request( + method=request.method, url=url, headers=headers, params=request.query_params, + content=await request.body(), + ) + return StreamingResponse( + resp.aiter_bytes(), status_code=resp.status_code, headers=dict(resp.headers) + ) + + +@app.api_route( + "/proxy/prometheus/{workspace}/{path:path}", + methods=["GET", "POST", "PUT", "DELETE"], +) +async def proxy_prometheus(workspace: str, path: str, request: Request): + url = f"{settings.controller_base_url}/v2/prometheus/{workspace}/{path}" + return await _proxy_request(request, url) + + +@app.api_route( + "/proxy/quickwit/{workspace}/{path:path}", + methods=["GET", "POST", "PUT", "DELETE"], +) +async def proxy_quickwit(workspace: str, path: str, request: Request): + url = f"{settings.controller_base_url}/v2/quickwit/{workspace}/{path}" + return await _proxy_request(request, url) + + +# ── Alert webhook (translate Grafana payload → controller schema) ──── + + +@app.post("/alerts/webhook") +async def alerts_webhook(workspace: str, request: Request): + def _translate_alert(alert: dict) -> dict: + _SEVERITY_CHOICES = {"info", "warning", "error", "success"} + _STATUS_TO_SEVERITY = {"firing": "warning", "resolved": "success"} + labels, annotations = alert.get("labels") or {}, alert.get("annotations") or {} + title = (annotations.get("summary") or labels.get("alertname") or "Grafana alert")[:255] + content = annotations.get("description") or title + severity = (labels.get("severity") if labels.get("severity") in _SEVERITY_CHOICES + else _STATUS_TO_SEVERITY.get(alert.get("status", "firing"), "info")) + message = {"category": "alert", "title": title, "content": content, "severity": severity} + generator_url = alert.get("generatorURL") + if generator_url: + message.update({"action_link": generator_url[:500], "action_text": "View in Grafana"}) + return message + + try: + token = await get_token() + except RuntimeError as e: + raise HTTPException(status_code=503, detail=str(e)) + grafana_payload = await request.json() + alerts = (grafana_payload or {}).get("alerts") or [] + translated = [_translate_alert(a) for a in alerts] + body = json.dumps({"workspace": workspace, "alerts": translated}) + resp = await http_client.post( + f"{settings.controller_base_url}/v2/alerts", + headers={"Authorization": f"Bearer {token}", "Content-Type": "application/json"}, + content=body, + ) + return Response(status_code=resp.status_code) + + +if __name__ == "__main__": + import uvicorn + uvicorn.run(app, host=args.bind, port=int(args.port)) diff --git a/rootfs/usr/share/grafana/oauth2/requirements.txt b/rootfs/usr/share/grafana/oauth2/requirements.txt new file mode 100644 index 0000000..11d9153 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/requirements.txt @@ -0,0 +1,10 @@ +fastapi>=0.68.0 +valkey>=6.1.0 +uvicorn[standard]>=0.15.0 +authlib>=1.0.0 +httpx>=0.23.0 +psycopg[binary]>=3.2.9 +python-jose>=3.3.0 +python-multipart>=0.0.5 +itsdangerous>=2.2.0 +pydantic_settings>=2.14.1 \ No newline at end of file diff --git a/rootfs/usr/share/grafana/oauth2/settings.py b/rootfs/usr/share/grafana/oauth2/settings.py new file mode 100644 index 0000000..8fab532 --- /dev/null +++ b/rootfs/usr/share/grafana/oauth2/settings.py @@ -0,0 +1,108 @@ +import re + +import valkey.asyncio +from pydantic import Field, computed_field, RedisDsn, HttpUrl +from pydantic_settings import BaseSettings, SettingsConfigDict + + +_DURATION_RE = re.compile(r"^\s*(\d+)\s*([smhd]?)\s*$", re.IGNORECASE) +_DURATION_MULTIPLIERS = {"": 1, "s": 1, "m": 60, "h": 3600, "d": 86400} + + +def _parse_duration(value: str) -> int: + """Parse a duration string like '60s', '5m', '1h', '7d' into seconds. + + A bare integer (no unit) is treated as seconds. Raises ValueError on + malformed input so misconfiguration fails fast at startup. + """ + if value is None: + raise ValueError("duration value must not be None") + match = _DURATION_RE.match(str(value)) + if not match: + raise ValueError(f"invalid duration: {value!r}") + number, unit = match.groups() + return int(number) * _DURATION_MULTIPLIERS[unit.lower()] + + +class Settings(BaseSettings): + model_config = SettingsConfigDict( + env_prefix="", + extra="ignore", + case_sensitive=True, + ) + + # Drycc configuration + drycc_valkey_url: RedisDsn = Field( + default="redis://localhost:6379/0", + validation_alias="DRYCC_VALKEY_URL", + ) + drycc_passport_url: HttpUrl = Field( + default="http://passport.drycc.cc", + validation_alias="DRYCC_PASSPORT_URL", + ) + drycc_passport_key: str = Field(default="", validation_alias="DRYCC_PASSPORT_KEY") + drycc_passport_secret: str = Field(default="", validation_alias="DRYCC_PASSPORT_SECRET") + drycc_passport_scopes: str = Field(default="", validation_alias="DRYCC_PASSPORT_SCOPES") + drycc_controller_url: HttpUrl = Field( + default="http://controller.drycc.cc", + validation_alias="DRYCC_CONTROLLER_URL", + ) + drycc_grafana_refresh: str = Field(default="60s", validation_alias="DRYCC_GRAFANA_REFRESH") + + drycc_state_check_cooldown: str = Field( + default="60s", + validation_alias="DRYCC_STATE_CHECK_COOLDOWN", + ) + drycc_session_ttl_fallback: str = Field( + default="7d", + validation_alias="DRYCC_SESSION_TTL_FALLBACK", + ) + + # Grafana configuration + gf_security_admin_user: str = Field(default="admin", validation_alias="GF_SECURITY_ADMIN_USER") + gf_security_admin_password: str = Field( + default="admin", + validation_alias="GF_SECURITY_ADMIN_PASSWORD", + ) + gf_server_http_port: int = Field(default=3000, validation_alias="GF_SERVER_HTTP_PORT") + gf_database_url: str = Field(default="", validation_alias="GF_DATABASE_URL") + + @computed_field + @property + def passport_token_url(self) -> str: + return f"{str(self.drycc_passport_url).rstrip('/')}/oauth/token/" + + @computed_field + @property + def passport_userinfo_url(self) -> str: + return f"{str(self.drycc_passport_url).rstrip('/')}/oauth/userinfo/" + + @computed_field + @property + def controller_base_url(self) -> str: + """Controller URL normalized without trailing slash. + + Pydantic's HttpUrl appends a trailing slash on serialization, which + causes double slashes when callers concatenate paths with a leading + slash (e.g. f"{url}/v2/..."). Use this property for any path joining. + """ + return str(self.drycc_controller_url).rstrip('/') + + @computed_field + @property + def state_check_cooldown_seconds(self) -> int: + return _parse_duration(self.drycc_state_check_cooldown) + + @computed_field + @property + def session_ttl_fallback_seconds(self) -> int: + return _parse_duration(self.drycc_session_ttl_fallback) + + async def get_valkey_client(self) -> valkey.asyncio.Valkey: + return valkey.asyncio.from_url( + str(self.drycc_valkey_url), + decode_responses=True, + ) + + +settings = Settings() diff --git a/rootfs/usr/share/grafana/provisioning/alerting/contactpoints.yaml b/rootfs/usr/share/grafana/provisioning/alerting/contactpoints.yaml new file mode 100644 index 0000000..e217bd0 --- /dev/null +++ b/rootfs/usr/share/grafana/provisioning/alerting/contactpoints.yaml @@ -0,0 +1,12 @@ +apiVersion: 1 + +contactPoints: + - orgId: 1 + name: controller-alerts + receivers: + - uid: controller_alerts + name: controller alerts webhook + type: webhook + settings: + url: http://localhost:4000/alerts/webhook?workspace=drycc + httpMethod: POST diff --git a/rootfs/usr/share/grafana/provisioning/alerting/drycc.yaml b/rootfs/usr/share/grafana/provisioning/alerting/drycc.yaml new file mode 100644 index 0000000..881813d --- /dev/null +++ b/rootfs/usr/share/grafana/provisioning/alerting/drycc.yaml @@ -0,0 +1,803 @@ +apiVersion: 1 +groups: + - orgId: 1 + name: middle + folder: drycc + interval: 2m + rules: + - uid: eeq3o5l9asa2od + title: pod restart + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: increase(kube_pod_container_status_restarts_total{pod!~".*job.*"}[30m]) > 0 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 2m + keepFiringFor: 2m + annotations: + description: "namespace: {{ $labels.namespace }}, pod: {{ $labels.pod }} restart." + summary: Pod restart alert + isPaused: false + notification_settings: + receiver: controller-alerts + - uid: deq6i5dvnlkw0e + title: pod cpu usage over 80% + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: (sum(rate(container_cpu_usage_seconds_total{image!=""}[2m])) by (pod,namespace)/sum(kube_pod_container_resource_limits{resource="cpu"}) by (pod,namespace) * 100) > 80 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 2m + annotations: + description: "namespace: {{ $labels.namespace }}, pod: {{ $labels.pod }} 's cpu usage over {{$value }}%." + summary: CPU Usage of Pod limit on has exceeded 80%. + isPaused: false + notification_settings: + receiver: controller-alerts + - uid: beq6l7fts9nnke + title: pod memory usage over 80% + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: (sum(container_memory_working_set_bytes{pod!="",container!=""}) by (pod, namespace)) / (sum(container_spec_memory_limit_bytes{pod!=""}) by (pod, namespace)) > 0.8 and (sum(container_memory_working_set_bytes{pod!="",container!=""}) by (pod, namespace)) / (sum (container_spec_memory_limit_bytes{pod!=""}) by (pod, namespace)) < 2 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 2m + keepFiringFor: 2m + annotations: + description: "namespace: {{ $labels.namespace }}, pod: {{ $labels.pod }} 's memory usage over {{ $value }}%." + summary: Memory Usage of Pod limit on has exceeded 80%. + isPaused: false + notification_settings: + receiver: controller-alerts + - uid: eeq6lm6biyfb4e + title: pod start fail + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: kube_pod_status_phase{pod!~".*job.*", phase=~"Failed|Unknown"} == 1 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 2m + annotations: + description: "namespace: {{ $labels.namespace }}, pod: {{ $labels.pod }} start fail." + summary: Pod start fail + isPaused: false + notification_settings: + receiver: controller-alerts + - uid: feq6m6yr27togc + title: job exec error + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: kube_job_status_failed{job="drycc-kube-state-metrics"} == 1 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 2m + annotations: + description: "namespace: {{ $labels.namespace }}, job: {{ $labels.job_name }} exec {{ $labels.reason }}." + summary: Job exec error + isPaused: false + notification_settings: + receiver: controller-alerts + - uid: beqhp6i29jx8gb + title: pvc usage over 80% + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: kubelet_volume_stats_used_bytes{job="kubernetes-nodes"} / kubelet_volume_stats_capacity_bytes{job="kubernetes-nodes"} > 0.8 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 2m + annotations: + description: pvc capacity {{$labels.namespace}} {{$labels.persistentvolumeclaim}} on {{$labels.topology_topolvm_io_node}} usage over {{$value }}%. + summary: Pvc usage over 80% + isPaused: false + notification_settings: + receiver: controller-alerts + - uid: deqhpblphb7k0c + title: mysql down + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: mysql_up != 1 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 2m + annotations: + description: MySQL {{$labels.namespace}} {{$labels.pod}} on {{$labels.node}} is not up. + summary: MySQL not up. + isPaused: false + notification_settings: + receiver: controller-alerts + - uid: feqhpe2zz56o0d + title: mysql group replcation down + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: mysql_perf_schema_replication_group_member_info{member_state!="ONLINE"} == 1 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 2m + annotations: + description: MySQL group replication {{$labels.namespace}} {{$labels.pod}} on {{$labels.node}} is not up. + summary: MySQL group replication not up. + isPaused: false + notification_settings: + receiver: controller-alerts + - orgId: 1 + name: high + folder: drycc + interval: 1m + rules: + - uid: beq6ljd99ukg0c + title: node down + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: up{job="drycc-node-exporter"} == 0 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 1m + annotations: + description: "node: {{ $labels.instance }} down." + summary: Node down + isPaused: false + notification_settings: + receiver: controller-alerts + - uid: eeq6lu5wz3eo0c + title: cluster node over pressure + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: kube_node_status_condition{condition=~"OutOfDisk|MemoryPressure|DiskPressure",status!="false"} == 1 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 1m + annotations: + description: "node: {{ $labels.node }} has {{ $labels.condition }}." + summary: Cluster node over pressure + isPaused: false + notification_settings: + receiver: controller-alerts + - uid: deq6lwic91slce + title: cluster node status error + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: kube_node_status_condition{condition="Ready",status!="true"} == 1 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 1m + annotations: + description: "node: {{ $labels.node }} status error." + summary: Cluster node status error + isPaused: false + notification_settings: + receiver: controller-alerts + - orgId: 1 + name: middle + folder: drycc + interval: 2m + rules: + - uid: beq6ptjdxwmpsc + title: node cpu usage over 80% + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: (1 - avg by (instance) (irate(node_cpu_seconds_total{job="drycc-node-exporter",mode="idle"}[5m]))) * 100 > 80 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 2m + annotations: + description: "instance: {{ $labels.instance }} cpu usage over {{ $value }}%" + summary: Node cpu high over 80 + isPaused: false + notification_settings: + receiver: controller-alerts + - uid: ceqhohsqqyr5sb + title: node memory usage over 80% + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: (1-(node_memory_MemAvailable_bytes{job="drycc-node-exporter"} / (node_memory_MemTotal_bytes{job="drycc-node-exporter"})))* 100 > 80 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 2m + annotations: + description: "instance: {{ $labels.instance }} memory usage over {{ $value }}%." + summary: Node memory over 80 + isPaused: false + notification_settings: + receiver: controller-alerts + - uid: aeqhp0wbcsb9cf + title: node systen disk usage over 80% + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: (1 - node_filesystem_avail_bytes{job="drycc-node-exporter",fstype=~"ext4|xfs"} / node_filesystem_size_bytes{job="drycc-node-exporter",fstype=~"ext4|xfs"}) * 100 > 80 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + for: 2m + annotations: + description: "instance: {{ $labels.instance }} disk usage over {{ $value }}%." + summary: Node systen disk usage over 80 + isPaused: false + notification_settings: + receiver: controller-alerts + - orgId: 1 + name: slow + folder: drycc + interval: 1d + rules: + - uid: feq6lfrsmsum8a + title: node certificate expiration + condition: C + data: + - refId: A + relativeTimeRange: + from: 600 + to: 0 + datasourceUid: prometheus_on_drycc + model: + editorMode: code + expr: ((x509_cert_not_after - time()) / 86400) < 15 + instant: true + intervalMs: 1000 + legendFormat: __auto + maxDataPoints: 43200 + range: false + refId: A + - refId: C + datasourceUid: __expr__ + model: + conditions: + - evaluator: + params: + - 0 + type: gt + operator: + type: and + query: + params: + - C + reducer: + params: [] + type: last + type: query + datasource: + type: __expr__ + uid: __expr__ + expression: A + intervalMs: 1000 + maxDataPoints: 43200 + refId: C + type: threshold + noDataState: OK + execErrState: KeepLast + annotations: + description: "node: {{ $labels.node}} cert: {{ $labels.filepath }} will expire in {{ $labels.value }} days." + summary: Node certificate expiration + isPaused: false + notification_settings: + receiver: controller-alerts diff --git a/rootfs/usr/share/grafana/provisioning/dashboards/drycc.yaml b/rootfs/usr/share/grafana/provisioning/dashboards/drycc.yaml new file mode 100644 index 0000000..aa68d28 --- /dev/null +++ b/rootfs/usr/share/grafana/provisioning/dashboards/drycc.yaml @@ -0,0 +1,10 @@ +apiVersion: 1 +providers: +- name: 'default' + orgId: 1 + type: file + options: + path: /usr/share/grafana/dashboards + folder: 'drycc' + allowUiUpdates: false + disableDeletion: true diff --git a/rootfs/usr/share/grafana/provisioning/datasources/drycc.yaml b/rootfs/usr/share/grafana/provisioning/datasources/drycc.yaml new file mode 100644 index 0000000..85983ee --- /dev/null +++ b/rootfs/usr/share/grafana/provisioning/datasources/drycc.yaml @@ -0,0 +1,25 @@ +apiVersion: 1 +datasources: +- name: Application Logs + type: quickwit-quickwit-datasource + access: proxy + orgId: 1 + uid: application_logs + url: http://localhost:4000/proxy/quickwit/drycc + jsonData: + index: logs-* + logMessageField: log + version: 1 + editable: false + +- name: Prometheus on Drycc + type: prometheus + access: proxy + orgId: 1 + uid: prometheus_on_drycc + url: http://localhost:4000/proxy/prometheus/drycc + isDefault: true + jsonData: + httpMethod: POST + version: 1 + editable: false diff --git a/rootfs/usr/share/grafana/provisioning/plugins/drycc.yaml b/rootfs/usr/share/grafana/provisioning/plugins/drycc.yaml new file mode 100644 index 0000000..55439cb --- /dev/null +++ b/rootfs/usr/share/grafana/provisioning/plugins/drycc.yaml @@ -0,0 +1 @@ +apiVersion: 1 diff --git a/telegraf/Makefile b/telegraf/Makefile deleted file mode 100644 index c58adc3..0000000 --- a/telegraf/Makefile +++ /dev/null @@ -1,4 +0,0 @@ -SHORT_NAME ?= telegraf -SHELL_SCRIPTS = rootfs/home/telegraf/start-telegraf - -include ../default.mk diff --git a/telegraf/README.md b/telegraf/README.md deleted file mode 100644 index e1ef924..0000000 --- a/telegraf/README.md +++ /dev/null @@ -1,47 +0,0 @@ -# [Telegraf](https://influxdata.com/time-series-platform/telegraf/) - -## Description -Telegraf is a metrics collection daemon from InfluxData. It contains numerous input and output plugins that allows the user to customize what data they collect and where it is sent. This image is based on the official [drycc base image](https://github.com/drycc/docker-base). - -## Configuration -Telegraf configuration is based largely on a toml file that is passed in when the binary starts. The issue with doing this in a containerized environment is how can you "dynamically" build this file based on values passed into the container at runtime. Therefore, this image relies on a project called [envtpl](https://github.com/arschles/envtpl) to produce the telegraf configuration file. It can take environment variables and through using go templates produce the necessary stanzas in the toml file to start telegraf. Currently, the go template only supports basic if checks and outputting values that have been set. - -## Host information -You must do 2 things if you want to receive host level metrics from telegraf. - * - Set the volume mounts for both `/sys` and `/proc` - * - Set the environment variables `HOST_PROC` and `HOST_SYS` to the values of where `/sys` and `/proc` are mounted in the container. Example entries can be found in the [manifest](manifests/drycc-monitor-telegraf-daemon.yaml). - - ## Environment Variables - The configuration is driven via environment variables which are published to the `config.toml` file passed to telegraf when it starts. The following table gives the environment variable name and the default value if it is not set. - - | Name | Default | Description | - |-----------|---------------|---------------| - | AGENT_INTERVAL | 10s | Default data collection interval for all inputs | - | AGENT_ROUND_INTERVAL | true | Rounds collection interval to 'interval' ie, if interval="10s" then always collect on :00, :10, :20, etc. | - | AGENT_BUFFER_LIMIT | 10000 | Telegraf will cache metric_buffer_limit metrics for each output, and will flush this buffer on a successful write. | - | AGENT_COLLECTION_JITTER | 0s | Collection jitter is used to jitter the collection by a random amount. Each plugin will sleep for a random time within jitter before collecting. This can be used to avoid many plugins querying things like sysfs at the same time, which can have a measurable effect on the system. | - | AGENT_FLUSH_INTERVAL | 10s | Default data flushing interval for all outputs. You should not set this below interval. Maximum flush_interval will be flush_interval + flush_jitter | - | AGENT_FLUSH_JITTER | 0s | Jitter the flush interval by a random amount. This is primarily to avoid large write spikes for users running a large number of telegraf instances. ie, a jitter of 5s and flush_interval 10s means flushes will happen every 10-15s. | - | AGENT_DEBUG | false | Run telegraf in debug mode. | - | AGENT_QUIET | false | Run telegraf in quiet mode. | - | AGENT_HOSTNAME | NodeName | Override default hostname | - - -## Development -The provided `Makefile` has various targets to help support building and publishing new images into a kubernetes cluster. - -### Environment variables -There are a few key environment variables you should be aware of when interacting with the `make` targets. - -* `BUILD_TAG` - The tag provided to the docker image when it is built (defaults to the git-sha) -* `SHORT_NAME` - The name of the image (defaults to `grafana`) -* `DRYCC_REGISTRY` - This is the registry you are using (default `dockerhub`) -* `IMAGE_PREFIX` - This is the account for the registry you are using (default `drycc`) - -### Make targets - -* `make build` - Build docker image -* `make push` - Push docker image to a registry -* `make upgrade` - Replaces the running grafana instance with a new one - -The typical workflow will look something like this - `DRYCC_REGISTRY= IMAGE_PREFIX=foouser make build push upgrade`` diff --git a/telegraf/rootfs/Dockerfile b/telegraf/rootfs/Dockerfile deleted file mode 100644 index 9095e5a..0000000 --- a/telegraf/rootfs/Dockerfile +++ /dev/null @@ -1,36 +0,0 @@ -FROM registry.drycc.cc/drycc/base:bullseye - -ENV DRYCC_UID=1001 \ - DRYCC_GID=1001 \ - DRYCC_HOME_DIR=/home/telegraf \ - TELEGRAF_VERSION="1.23.4" \ - ENVTPL_VERSION="1.0.0" \ - JQ_VERSION="1.6" - -RUN groupadd drycc --gid ${DRYCC_GID} \ - && useradd drycc -u ${DRYCC_UID} -g ${DRYCC_GID} -s /bin/bash -m -d ${DRYCC_HOME_DIR} - -COPY . / - -RUN install-stack telegraf ${TELEGRAF_VERSION} \ - && install-stack envtpl ${ENVTPL_VERSION} \ - && install-stack jq ${JQ_VERSION} \ - && rm -rf \ - /usr/share/doc \ - /usr/share/man \ - /usr/share/info \ - /usr/share/locale \ - /var/lib/apt/lists/* \ - /var/log/* \ - /var/cache/debconf/* \ - /etc/systemd \ - /lib/lsb \ - /lib/udev \ - /usr/lib/`echo $(uname -m)`-linux-gnu/gconv/IBM* \ - /usr/lib/`echo $(uname -m)`-linux-gnu/gconv/EBC* \ - && mkdir -p /usr/share/man/man{1..8} \ - && chown -R ${DRYCC_UID}:${DRYCC_GID} ${DRYCC_HOME_DIR} - -USER ${DRYCC_UID} -WORKDIR ${DRYCC_HOME_DIR} -CMD ["/home/telegraf/start-telegraf"] diff --git a/telegraf/rootfs/home/telegraf/config.toml.tpl b/telegraf/rootfs/home/telegraf/config.toml.tpl deleted file mode 100644 index db59563..0000000 --- a/telegraf/rootfs/home/telegraf/config.toml.tpl +++ /dev/null @@ -1,288 +0,0 @@ -# Set Tag Configuration -[tags] - -{{- if .GLOBAL_TAGS }} -[global_tags] - {{ range $index, $item := split "," .GLOBAL_TAGS }} - {{ $value := split ":" $item }}{{ $value._0 }}={{ $value._1 | quote }} - {{ end }} -{{ end }} - -# Set Agent Configuration -[agent] - interval = {{ default "10s" .AGENT_INTERVAL | quote }} - round_interval = {{ default true .AGENT_ROUND_INTERVAL }} - metric_buffer_limit = {{ default "10000" .AGENT_BUFFER_LIMIT }} - collection_jitter = {{ default "1s" .AGENT_COLLECTION_JITTER | quote }} - flush_interval = {{ default "1s" .AGENT_FLUSH_INTERVAL | quote }} - flush_jitter = {{ default "0s" .AGENT_FLUSH_JITTER | quote }} - debug = {{ default false .AGENT_DEBUG }} - quiet = {{ default false .AGENT_QUIET }} - flush_buffer_when_full = {{ default true .AGENT_FLUSH_BUFFER }} - {{- if .AGENT_HOSTNAME }}hostname = {{ .AGENT_HOSTNAME | quote }} {{ end }} - -# Set output configuration -{{- if .AMON_INSTANCE }} -[[outputs.amon]] - server_key = {{ .AMON_SERVER_KEY | quote }} - amon_instance = {{ .AMON_INSTANCE | quote }} - timeout = {{ default "5s" .AMON_TIMEOUT | quote }} -{{ end }} - -{{- if .AMQP_URL }} -[[outputs.amqp]] - url = {{ .AMQP_URL | quote }} - exchange = {{ default "telegraf" .AMQP_EXCHANGE | quote }} - routing_tag = {{ default "host" .AMQP_ROUTING_TAG | quote }} - {{- if .AMQP_SSL_CA }} ssl_ca = {{ .AMQP_SSL_CA | quote }} {{ end }} - {{- if .AMQP_SSL_CERT }} ssl_cert = {{ .AMQP_SSL_CERT | quote }} {{ end }} - {{- if .AMQP_SSL_KEY }} ssl_key = {{ .AMQP_SSL_KEY | quote }} {{ end }} - retention_policy = {{ default "default" .AMQP_RETENTION_POLICY | quote }} - database = = {{ default "telegraf" .AMQP_DATABASE | quote }} - precision = {{ default "s" .AMQP_PRECISION | quote }} -{{ end }} - -{{- if .DATADOG_API_KEY }} -[[outputs.datadog]] - apikey = {{ .DATADOG_API_KEY | quote }} -{{ end }} - -{{- if .GRAPHITE_SERVERS }} -[[outputs.graphite]] - servers = [{{ .GRAPHITE_SERVERS }}] - prefix = {{ default "" .GRAPHITE_PREFIX | quote }} - timeout = {{ default 2 .GRAPHITE_TIMEOUT }} -{{ end }} - -{{- if .INFLUXDB_URLS }} -[[outputs.influxdb]] - urls = [{{ .INFLUXDB_URLS }}] - database = {{default "kubernetes" .INFLUXDB_DATABASE | quote }} - precision = {{ default "ns" .INFLUXDB_PRECISION | quote }} - timeout = {{ default "5s" .INFLUXDB_TIMEOUT | quote }} - {{- if .INFLUXDB_USER}} username = {{ .INFLUXDB_USER | quote }} {{ end }} - {{- if .INFLUXDB_PASSWORD}} password = {{ .INFLUXDB_PASSWORD | quote }} {{ end }} - {{- if .INFLUXDB_USER_AGENT}} user_agent = {{ .INFLUXDB_USER_AGENT | quote }} {{ end }} - {{- if .INFLUXDB_UDP_PAYLOAD}} udp_payload = {{ .INFLUXDB_UDP_PAYLOAD | quote }} {{ end }} -{{ end }} - -{{- if .INFLUXDB_URLS_V2}} -[[outputs.influxdb_v2]] - urls = [{{ .INFLUXDB_URLS_V2 }}] - bucket = {{default "kubernetes" .INFLUXDB_BUCKET | quote }} - organization = {{default "drycc" .INFLUXDB_ORG | quote }} - token = {{default "" .INFLUXDB_TOKEN | quote }} -{{ end }} - -{{- if .KAFKA_BROKERS}} -[[outputs.kafka]] - brokers = [{{ .KAFKA_BROKERS }}] - topic = {{ default "telegraf" .KAFKA_TOPIC | quote }} - routing_tag = {{ .KAFKA_ROUTING_TAG | quote }} - {{- if .KAFKA_CERTIFICATE }} certificate = {{ .KAFKA_CERTIFICATE | quote }} {{ end }} - {{- if .KAFKA_KEY }} key = {{ .KAFKA_KEY | quote }} {{ end }} - {{- if .KAFKA_CA }} ca = {{ .KAFKA_CA | quote }} {{ end }} - {{- if .KAFKA_VERIFY_SSL }} verify_ssl = {{ .KAFKA_VERIFY_SSL }} {{ end }} -{{ end }} - -{{- if .LIBRATO_API_TOKEN }} -[[outputs.librato]] - api_user = {{ .LIBRATO_API_USER | quote }} - api_token = {{ .LIBRATO_API_TOKEN | quote }} - source_tag = {{ .LIBRATO_SOURCE_TAG | quote }} -{{ end }} - -{{- if .OPEN_TSDB_HOST }} - prefix = {{ .OPEN_TSDB_PREFIX | quote }} - host = {{ .OPEN_TSDB_HOST | quote }} - port = {{ .OPEN_TSDB_PORT }} - debug = {{ .OPEN_TSDB_DEUBG }} -{{ end }} - -{{- if .RIEMANN_URL }} -[[outputs.riemann]] - url = {{ .RIEMANN_URL | quote }} - transport = {{ .RIEMANN_TRANSPORT | quote }} -{{ end }} - -# Set Input Configuration -{{- if .ENABLE_BARE_METAL_SERVER }} -[[inputs.netstat]] -[[inputs.swap]] -[[inputs.system]] -[[inputs.mem]] -[[inputs.cpu]] - percpu = true - totalcpu = true -[[inputs.diskio]] - {{- if .DISKIO_DEVICES }} devices = [{{ .DISKIO_DEVICES }}] {{ end }} - {{- if .DISKIO_SKIP_SERIAL_NUMBER }} skip_serial_number = {{ .DISKIO_SKIP_SERIAL_NUMBER }} {{ end }} - -[[inputs.net]] - {{- if .NET_INTERFACES }} interfaces = [{{ .NET_INTERFACES}}] {{ end }} -[[inputs.disk]] - {{- if .DISK_MOUNT_POINTS }} mount_points = [{{ .DISK_MOUNT_POINTS }}] {{ end }} -{{ end }} - -{{- if .ENABLE_MONITORSELF }}[[inputs.internal]]{{ end }} -{{- if .ENABLE_ZFS }}[[inputs.zfs]]{{ end }} - -{{- if .APACHE_URLS }} -[[inputs.apache]] - servers = [{{ .APACHE_URLS }}] -{{ end }} - -{{- if .DISQUE_SERVERS }} -[[inputs.disque]] - servers = [{{ .DISQUE_SERVERS }}] -{{ end }} - -{{- if .DOCKER_ENDPOINT }} -[[inputs.docker]] - endpoint = {{ .DOCKER_ENDPOINT | quote }} - container_names = [{{ if .DOCKER_CONTAINER_NAMES }}{{ .DOCKER_CONTAINER_NAMES }}{{ end }}] -{{ end }} - -{{- if .ELASTIC_SEARCH_SERVERS }} -[[inputs.elasticsearch]] - servers = [{{ .ELASTIC_SEARCH_SERVERS }}] - local = {{ .ELASTIC_SEARCH_LOCAL }} - cluster_health = {{ .ELASTIC_SEARCH_CLUSTER_HEALTH }} -{{ end }} - -{{- if .HAPROXY_SERVERS }} -[[inputs.haproxy]] - servers = [{{ .HAPROXY_SERVERS }}] -{{ end }} - -{{- if .HTTP_JSON_SERVERS }} -[[inputs.httpjson]] - name = {{ .HTTP_JSON_NAME | quote }} - servers = [{{ .HTTP_JSON_SERVERS }}] - method = {{ .HTTP_JSON_METHOD | quote }} - tag_keys = [{{ .HTTP_JSON_TAG_KEYS }}] - [json.parameters] - event_type = {{ .HTTP_JSON_EVENT_TYPE | quote }} - threshold = {{ .HTTP_JSON_THRESHOLD | quote }} -{{ end }} - -{{- if .ENABLE_INFLUXDB_INPUT }} -[[inputs.influxdb]] - urls = [{{ .INFLUXDB_INPUT_URLS }}] -{{ end }} - -{{- if .ENABLE_INFLUXDB_V2_INPUT }} -[[inputs.prometheus]] - urls = [{{ .INFLUXDB_V2_INPUT_URLS }}] -{{ end }} - -{{- if and .KUBERNETES_URL .ENABLE_KUBERNETES }} -[[inputs.kubernetes]] - url = {{ .KUBERNETES_URL | quote }} - bearer_token = {{ .KUBERNETES_BEARER_TOKEN_PATH | quote }} - {{- if .KUBERNETES_SSL_CA }}tls_ca = {{ .KUBERNETES_SSL_CA | quote }} {{ end }} - {{- if .KUBERNETES_SSL_CERT }}tls_cert = {{ .KUBERNETES_SSL_CERT | quote }} {{ end }} - {{- if .KUBERNETES_SSL_KEY }}tls_key = {{ .KUBERNETES_SSL_KEY | quote }} {{ end }} - insecure_skip_verify = {{ default true .KUBERNETES_INSECURE_SKIP_VERIFY }} -{{ end }} - -{{- if and .KUBE_INVENTORY_URL .ENABLE_KUBE_INVENTORY }} -[[inputs.kube_inventory]] - url = {{ .KUBE_INVENTORY_URL | quote }} - bearer_token = {{ .KUBERNETES_BEARER_TOKEN_PATH | quote }} - {{- if .KUBERNETES_SSL_CA }}tls_ca = {{ .KUBERNETES_SSL_CA | quote }} {{ end }} - {{- if .KUBERNETES_SSL_CERT }}tls_cert = {{ .KUBERNETES_SSL_CERT | quote }} {{ end }} - {{- if .KUBERNETES_SSL_KEY }}tls_key = {{ .KUBERNETES_SSL_KEY | quote }} {{ end }} - {{- if .KUBE_RESOURCE_INCLUE }}resource_include = {{ .KUBE_RESOURCE_INCLUE }} {{ end }} - {{- if .KUBE_RESOURCE_EXCLUE }}resource_exclude = {{ .KUBE_RESOURCE_EXCLUE }} {{ end }} - {{- if .KUBE_SELECTOR_INCLUE }}selector_include = {{ .KUBE_SELECTOR_INCLUE }} {{ end }} - {{- if .KUBE_SELECTOR_EXCLUE }}selector_exclude = {{ .KUBE_SELECTOR_EXCLUE }} {{ end }} - insecure_skip_verify = {{ default true .KUBERNETES_INSECURE_SKIP_VERIFY }} -{{ end }} - -{{- if .MEMCACHED_SERVERS }} -[[inputs.memcached]] - servers = [{{ .MEMCACHED_SERVERS }}] -{{ end }} - -{{- if .MONGODB_SERVERS }} -[[inputs.mongodb]] - servers = [{{ .MONGODB_SERVERS }}] -{{ end }} - -{{- if .MYSQL_SERVERS }} -[[inputs.mysql]] - servers = [{{ .MYSQL_SERVERS }}] -{{ end }} - -{{- if .NGINX_URLS }} -[[inputs.nginx]] - urls = [{{ .NGINX_URLS }}] -{{ end }} - -{{- if .POSTGRESQL_ADDRESS }} -[[inputs.postgresql]] - address = {{ .POSTGRESQL_ADDRESS | quote }} - {{- if .POSTGRESQL_DATABASES }} databases = [{{ .POSTGRESQL_DATABASES }}] {{ end }} -{{ end }} - -{{- if and .PROMETHEUS_URLS .ENABLE_PROMETHEUS }} -[[inputs.prometheus]] - urls = [{{ .PROMETHEUS_URLS }}] - insecure_skip_verify = {{ default true .PROMETHEUS_INSECURE_SKIP_VERIFY }} - bearer_token = {{ .PROMETHEUS_BEARER_TOKEN_PATH | quote }} -{{ end }} - -{{- if .ENABLE_ETCD }} -[[inputs.prometheus]] - urls = [{{ .ETCD_URLS }}] -{{ end }} - -{{- if .RABBITMQ_URL }} -[[inputs.rabbitmq]] - url = {{ .RABBITMQ_URL | quote }} - {{- if .RABBITMQ_NAME}} name = {{ .RABBITMQ_NAME | quote }} {{ end }} - {{- if .RABBITMQ_USERNAME }} username = {{ .RABBITMQ_USERNAME | quote }} {{ end }} - {{- if .RABBITMQ_PASSWORD }} password = {{ .RABBITMQ_PASSWORD | quote }} {{ end }} - {{- if .RABBITMQ_NODES }} nodes = [{{ .RABBITMQ_NODES}}] {{ end }} -{{ end }} - -{{- if .REDIS_SERVERS }} -[[inputs.redis]] - servers = {{ .REDIS_SERVERS }} -{{ end }} - -{{- if .RETHINKDB_SERVERS }} -[[inputs.rethinkdb]] - servers = [{{ .RETHINKDB_SERVERS }}] -{{ end }} - -{{- if .ZOOKEEPER_SERVERS }} -[[inputs.zookeeper]] - servers = [{{ .ZOOKEEPER_SERVERS }}] -{{ end }} - -# Set Service Input Configuration -{{- if .KAFKA_CONSUMER_TOPICS }} -[[inputs.kafka_consumer]] - topics = [{{ .KAFKA_CONSUMER_TOPICS }}] - zookeeper_peers = [{{ .KAFKA_CONSUMER_ZOOKEEPER_PEERS }}] - consumer_group = {{ .KAFKA_CONSUMER_GROUP | quote }} - point_buffer = {{ .KAFKA_CONSUMER_POINT_BUFFER }} - offset = {{ .KAFKA_CONSUMER_OFFSET | quote }} -{{ end }} - -{{- if .STATSD_SERVICE_ADDRESS }} -[[inputs.statsd]] - service_address = {{ .STATSD_SERVICE_ADDRESS | quote }} - delete_gauges = {{ default false .STATSD_DELETE_GAUGES }} - delete_counters = {{ default false .STATSD_DELETE_COUNTERS }} - delete_sets = {{ default false .STATSD_DELETE_SETS }} - delete_timings = {{ default true .STATSD_DELETE_TIMINGS}} - percentiles = [{{ default 90 .STATSD_PERCENTILES }}] - convert_names = {{ default true .STATSD_CONVERT_NAMES }} - {{- if .STATSD_TEMPLATES}} templates = [{{ .STATSD_TEMPLATES }}] {{ end }} - allowed_pending_messages = {{ default 10000 .STATSD_ALLOWED_PENDING_MESSAGES }} - percentile_limit = {{ default 1000 .STATSD_PERCENTILE_LIMIT }} - udp_packet_size = {{ default 1500 .STATSD_UDP_PACKET_SIZE }} -{{ end }} diff --git a/telegraf/rootfs/home/telegraf/start-telegraf b/telegraf/rootfs/home/telegraf/start-telegraf deleted file mode 100755 index 755ec21..0000000 --- a/telegraf/rootfs/home/telegraf/start-telegraf +++ /dev/null @@ -1,52 +0,0 @@ -#!/bin/bash -set -e - -API_URL=${API_URL:-"https://$KUBERNETES_SERVICE_HOST:$KUBERNETES_SERVICE_PORT"} - -export BEARER_TOKEN_PATH TOKEN NODE_NAME NODE_IP AGENT_HOSTNAME -BEARER_TOKEN_PATH=/var/run/secrets/kubernetes.io/serviceaccount/token -TOKEN="$(cat $BEARER_TOKEN_PATH)" -NODE_NAME="$(curl -s "$API_URL/api/v1/namespaces/$POD_NAMESPACE/pods/$HOSTNAME" --header "Authorization: Bearer $TOKEN" --insecure | jq -r '.spec.nodeName')" -echo "Node Name set (${NODE_NAME})" -NODE_IP="$(curl -s "$API_URL/api/v1/namespaces/$POD_NAMESPACE/pods/$HOSTNAME" --header "Authorization: Bearer $TOKEN" --insecure | jq -r '.status.hostIP')" -echo "Node IP set (${NODE_IP})" -AGENT_HOSTNAME=$NODE_NAME - -if [ -n "$ENABLE_KUBERNETES" ]; then - export KUBERNETES_BEARER_TOKEN_PATH=${BEARER_TOKEN_PATH} - export KUBERNETES_URL="https://${NODE_IP}:10250" - echo "Setting KUBERNETES_URL: $KUBERNETES_URL" -fi - -if [ -n "$ENABLE_PROMETHEUS" ]; then - export PROMETHEUS_BEARER_TOKEN_PATH=${BEARER_TOKEN_PATH} - export PROMETHEUS_URLS="\"$API_URL/api/v1/proxy/nodes/$NODE_NAME/metrics\", \"$API_URL/metrics\"" - echo "Setting PROMETHEUS_URLS: $PROMETHEUS_URLS" -fi - -# if the influxdb url does not start with a quote, assume it's singular and quote it -# shellcheck disable=SC2153 -if [ -n "$INFLUXDB_URL" ] && [ "${INFLUXDB_URLS:0:1}" != "\"" ]; then - export INFLUXDB_URLS="\"$INFLUXDB_URLS\"" -fi -# redis servers -if [ -n "$DRYCC_REDIS_ADDRS" ]; then - REDIS_SERVERS="" - for ADDR in $(echo "${DRYCC_REDIS_ADDRS}" | tr "," "\n") - do - REDIS_SERVERS="\":$DRYCC_REDIS_PASSWORD@$ADDR\", ${REDIS_SERVERS}" - done - REDIS_SERVERS=\["${REDIS_SERVERS::-2}"\] - export REDIS_SERVERS -fi - -echo "Building config.toml!" -envtpl > config.toml < config.toml.tpl | sed '/^$/d' - -echo "Finished building toml..." -echo "###########################################" -echo "###########################################" -cat config.toml -echo "###########################################" -echo "###########################################" -telegraf -config config.toml -quiet diff --git a/versioning.mk b/versioning.mk index 02ebe33..4396eab 100644 --- a/versioning.mk +++ b/versioning.mk @@ -10,13 +10,13 @@ info: @echo "Immutable tag: ${IMAGE}" @echo "Mutable tag: ${MUTABLE_IMAGE}" -.PHONY: docker-push -docker-push: docker-mutable-push docker-immutable-push +.PHONY: podman-push +podman-push: podman-mutable-push podman-immutable-push -.PHONY: docker-immutable-push -docker-immutable-push: - docker push ${IMAGE} +.PHONY: podman-immutable-push +podman-immutable-push: + podman push ${IMAGE} -.PHONY: docker-mutable-push -docker-mutable-push: - docker push ${MUTABLE_IMAGE} +.PHONY: podman-mutable-push +podman-mutable-push: + podman push ${MUTABLE_IMAGE}