chore(controller): remove cluster domain

Author: duanhongyi

charts/controller/templates/_helpers.tpl

@@ -17,8 +17,6 @@ env:
   value: "{{ (tpl .Values.filerImage .) }}"
 - name: "DRYCC_FILER_IMAGE_PULL_POLICY"
   value: "{{ (tpl .Values.filerImagePullPolicy .) }}"
-- name: "KUBERNETES_CLUSTER_DOMAIN"
-  value: "{{ .Values.global.clusterDomain }}"
 - name: "DRYCC_APP_GATEWAY_CLASS"
   value: "{{ .Values.appGatewayClass }}"
 {{- if (.Values.appStorageClass) }}
@@ -69,7 +67,7 @@ env:
       name: valkey-creds
       key: password
 - name: DRYCC_VALKEY_URL
-  value: "redis://:$(VALKEY_PASSWORD)@drycc-valkey.{{.Release.Namespace}}.svc.{{.Values.global.clusterDomain}}:16379/0"
+  value: "redis://:$(VALKEY_PASSWORD)@drycc-valkey:16379/0"
 {{- end }}
 {{- if (.Values.databaseReplicaUrl) }}
 - name: DRYCC_DATABASE_REPLICA_URL
@@ -96,9 +94,9 @@ env:
       name: database-creds
       key: password
 - name: DRYCC_DATABASE_URL
-  value: "postgres://$(DRYCC_PG_USER):$(DRYCC_PG_PASSWORD)@drycc-database.{{.Release.Namespace}}.svc.{{.Values.global.clusterDomain}}:5432/controller"
+  value: "postgres://$(DRYCC_PG_USER):$(DRYCC_PG_PASSWORD)@drycc-database:5432/controller"
 - name: DRYCC_DATABASE_REPLICA_URL
-  value: "postgres://$(DRYCC_PG_USER):$(DRYCC_PG_PASSWORD)@drycc-database-replica.{{.Release.Namespace}}.svc.{{.Values.global.clusterDomain}}:5432/controller"
+  value: "postgres://$(DRYCC_PG_USER):$(DRYCC_PG_PASSWORD)@drycc-database-replica:5432/controller"
 {{- end }}
 {{- if (.Values.workflowManagerUrl) }}
 - name: WORKFLOW_MANAGER_URL
@@ -120,7 +118,7 @@ env:
       key: victoriametrics-url
 {{- else if .Values.victoriametrics.enabled }}
 - name: "DRYCC_VICTORIAMETRICS_URL"
-  value: "http://drycc-victoriametrics-vmselect.{{$.Release.Namespace}}.svc.{{$.Values.global.clusterDomain}}:8481"
+  value: "http://drycc-victoriametrics-vmselect:8481"
 {{- end }}
 {{- if .Values.passport.enabled }}
 - name: "DRYCC_PASSPORT_URL"
@@ -157,7 +155,7 @@ env:
       key: passport-secret
 {{- end }}
 - name: QUICKWIT_SEARCHER_URL
-  value: http://drycc-quickwit-searcher.{{ $.Release.Namespace }}.svc.{{ .Values.global.clusterDomain }}:7280
+  value: http://drycc-quickwit-searcher:7280
 - name: QUICKWIT_LOG_INDEX_PREFIX
   value: {{ .Values.quickwit.logIndexPrefix }}
 {{- range $key, $value := .Values.environment }}

charts/controller/templates/controller-celery-deloyment.yaml

@@ -34,9 +34,7 @@ spec:
           - netcat
           - -v
           - -u
-          - $(DRYCC_DATABASE_URL),$(DRYCC_VALKEY_URL),$(DRYCC_DATABASE_REPLICA_URL)
-          - -a
-          - $(DRYCC_CONTROLLER_API_SERVICE_HOST):$(DRYCC_CONTROLLER_API_SERVICE_PORT)
+          - $(DRYCC_DATABASE_URL),$(DRYCC_VALKEY_URL),$(DRYCC_DATABASE_REPLICA_URL),http://drycc-controller-api
         {{- include "controller.envs" . | indent 8 }}
       containers:
       - name: drycc-controller-celery

charts/controller/templates/controller-cronjob-daily.yaml

@@ -29,9 +29,7 @@ spec:
               - netcat
               - -v
               - -u
-              - $(DRYCC_DATABASE_URL),$(DRYCC_VALKEY_URL),$(DRYCC_DATABASE_REPLICA_URL)
-              - -a
-              - $(DRYCC_CONTROLLER_API_SERVICE_HOST):$(DRYCC_CONTROLLER_API_SERVICE_PORT)
+              - $(DRYCC_DATABASE_URL),$(DRYCC_VALKEY_URL),$(DRYCC_DATABASE_REPLICA_URL),http://drycc-controller-api
             {{- include "controller-job.envs" . | indent 12 }}
           containers:
           - image: {{.Values.imageRegistry}}/{{.Values.imageOrg}}/controller:{{.Values.imageTag}}

charts/controller/templates/controller-cronjob-hourly.yaml

@@ -30,8 +30,8 @@ spec:
               - -v
               - -u
               - $(DRYCC_DATABASE_URL),$(DRYCC_VALKEY_URL),$(DRYCC_DATABASE_REPLICA_URL)
-              - -a
-              - $(DRYCC_CONTROLLER_API_SERVICE_HOST):$(DRYCC_CONTROLLER_API_SERVICE_PORT)
+              - -u
+              - $(DRYCC_DATABASE_URL),$(DRYCC_VALKEY_URL),$(DRYCC_DATABASE_REPLICA_URL),http://drycc-controller-api
             {{- include "controller-job.envs" . | indent 12 }}
           containers:
           - image: {{.Values.imageRegistry}}/{{.Values.imageOrg}}/controller:{{.Values.imageTag}}

charts/controller/templates/controller-mutate-deloyment.yaml

@@ -24,9 +24,7 @@ spec:
         - netcat
         - -v
         - -u
-        - $(DRYCC_DATABASE_URL),$(DRYCC_VALKEY_URL),$(DRYCC_DATABASE_REPLICA_URL)
-        - -a
-        - $(DRYCC_CONTROLLER_API_SERVICE_HOST):$(DRYCC_CONTROLLER_API_SERVICE_PORT)
+        - $(DRYCC_DATABASE_URL),$(DRYCC_VALKEY_URL),$(DRYCC_DATABASE_REPLICA_URL),http://drycc-controller-api
         {{- include "controller.envs" . | indent 8 }}
       containers:
       - name: drycc-controller

charts/controller/values.yaml

@@ -178,11 +178,6 @@ victoriametrics:
 global:
   # Admin email, used for each component to send email to administrator
   email: "drycc@drycc.cc"
-  # 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"

rootfs/api/models/service.py

@@ -1,7 +1,6 @@
 import logging
 from functools import partial
 from django.db import models
-from django.conf import settings
 from django.contrib.auth import get_user_model
 from api.utils import validate_json
 from api.exceptions import ServiceUnavailable
@@ -53,9 +52,7 @@ def name(self):
 
     @property
     def domain(self):
-        return "{}.{}.svc.{}".format(
-            self.name, self.namespace, settings.KUBERNETES_CLUSTER_DOMAIN
-        )
+        return "{}.{}.svc".format(self.name, self.namespace)
 
     @property
     def namespace(self):

rootfs/api/settings/production.py

@@ -266,18 +266,11 @@ def randstr(k):
 # the k8s namespace in which the controller and workflow were installed.
 WORKFLOW_NAMESPACE = os.environ.get('WORKFLOW_NAMESPACE', 'drycc')
 
-# kubernetes cluster domain
-KUBERNETES_CLUSTER_DOMAIN = os.environ.get("KUBERNETES_CLUSTER_DOMAIN", "cluster.local")
 # default scheduler settings
 SCHEDULER_MODULE = 'scheduler'
 SCHEDULER_URL = "https://{}:{}".format(
-    # accessing the k8s api server by IP address rather than hostname avoids
-    # intermittent DNS errors
-    os.environ.get(
-        'KUBERNETES_SERVICE_HOST',
-        'kubernetes.default.svc.{}'.format(KUBERNETES_CLUSTER_DOMAIN)
-    ),
-    os.environ.get('KUBERNETES_SERVICE_PORT', '443')
+    os.environ.get('KUBERNETES_SERVICE_HOST', 'kubernetes.default'),
+    os.environ.get('KUBERNETES_SERVICE_PORT', '443'),
 )
 
 K8S_API_VERIFY_TLS = os.environ.get('K8S_API_VERIFY_TLS', 'true').lower() == "true"

rootfs/api/tests/test_app.py

@@ -452,8 +452,7 @@ def test_app_verify_application_health_success(self, mock_requests):
 
         # create app
         app_id = self.create_app()
-        hostname = 'http://{}.{}.svc.{}:80/'.format(
-            app_id, app_id, settings.KUBERNETES_CLUSTER_DOMAIN)
+        hostname = 'http://{}.{}.svc:80/'.format(app_id, app_id)
         mr = mock_requests.register_uri('GET', hostname, responses)
 
         # deploy app to get verification
@@ -488,8 +487,7 @@ def test_app_verify_application_health_failure_404(self, mock_requests):
         # create app
         app_id = self.create_app()
 
-        hostname = 'http://{}.{}.svc.{}:80/'.format(
-            app_id, app_id, settings.KUBERNETES_CLUSTER_DOMAIN)
+        hostname = 'http://{}.{}.svc:80/'.format(app_id, app_id)
         mr = mock_requests.register_uri('GET', hostname, responses)
         # deploy app to get verification
         url = "/v2/apps/{}/build".format(app_id)
@@ -512,8 +510,7 @@ def _raise_exception(request, ctx):
         # create app
         app_id = self.create_app()
         # function tries to hit router 10 times
-        hostname = 'http://{}.{}.svc.{}:80/'.format(
-            app_id, app_id, settings.KUBERNETES_CLUSTER_DOMAIN)
+        hostname = 'http://{}.{}.svc:80/'.format(app_id, app_id)
         mr = mock_requests.register_uri('GET', hostname, text=_raise_exception)
 
         # deploy app to get verification

rootfs/api/tests/test_service.py

@@ -1,6 +1,5 @@
 from django.contrib.auth import get_user_model
 from django.core.cache import cache
-from django.conf import settings
 
 from api.tests import DryccTransactionTestCase
 from api.tests.test_gateway import RouteTest
@@ -50,8 +49,7 @@ def test_service_basic_ops(self):
         self.assertEqual(response.status_code, 200, response.data)
         self.assertEqual(len(response.data['services']), 1)
         expected0 = {
-            "domain": "%s-%s.%s.svc.%s" % (
-                app_id, "test", app_id, settings.KUBERNETES_CLUSTER_DOMAIN),
+            "domain": "%s-%s.%s.svc" % (app_id, "test", app_id),
             "ports": [{
                 'name': "%s-%s-%s-%s" % (app_id, "test", 'udp', 5000),
                 'port': 5000,
@@ -75,8 +73,7 @@ def test_service_basic_ops(self):
 
         # add new port
         expected1 = {
-            "domain": "%s-%s.%s.svc.%s" % (
-                app_id, "test", app_id, settings.KUBERNETES_CLUSTER_DOMAIN),
+            "domain": "%s-%s.%s.svc" % (app_id, "test", app_id),
             "ports": [
                 {
                     'name': "%s-%s-%s-%s" % (app_id, "test", 'udp', 5000),
@@ -124,8 +121,7 @@ def test_service_basic_ops(self):
         self.assertEqual(response.status_code, 200, response.data)
         self.assertEqual(len(response.data['services']), 2)
         expected2 = {
-            "domain": "%s-%s.%s.svc.%s" % (
-                app_id, "test2", app_id, settings.KUBERNETES_CLUSTER_DOMAIN),
+            "domain": "%s-%s.%s.svc" % (app_id, "test2", app_id),
             "ports": [{
                 'name': "%s-%s-%s-%s" % (app_id, "test2", 'udp', 5000),
                 'port': 5000,