fix(services): prevent units from stopping as "failed/failed"

Deis' systemd units always return "failed/failed" after being stopped
by fleetctl. Somehow our ExecStop command to remove the service container
conflicts with our ExecStart command and causes it to return an error code.
This is fixed by removing the ExecStop commands and adding a `--rm` flag
to the docker command in ExecStart. Additionally, our on_exit() handler in
each /app/bin/boot script was not returning 0.

TESTING: Run `make run`, then `make stop` on a Deis cluster. You should see
all units stop and return to "inactive/dead" status, whereas before they
would show "failed/failed."

Author: mboersma

bin/boot

@@ -66,6 +66,7 @@ GUNICORN_PID=$!
 function on_exit() {
 	kill -TERM $CELERY_PID $GUNICORN_PID
 	wait $CELERY_PID $GUNICORN_PID 2>/dev/null
+	exit 0
 }
 trap on_exit INT TERM
 

systemd/deis-controller.service

@@ -8,8 +8,7 @@ EnvironmentFile=/etc/environment
 TimeoutStartSec=20m
 ExecStartPre=/bin/sh -c "docker history deis/controller >/dev/null || docker pull deis/controller:latest"
 ExecStartPre=/bin/sh -c "docker inspect deis-controller >/dev/null && docker rm -f deis-controller || true"
-ExecStart=/usr/bin/docker run --name deis-controller -p 8000:8000 -e PUBLISH=8000 -e HOST=${COREOS_PRIVATE_IPV4} --volumes-from=deis-logger deis/controller
-ExecStop=/usr/bin/docker rm -f deis-controller
+ExecStart=/usr/bin/docker run --name deis-controller --rm -p 8000:8000 -e PUBLISH=8000 -e HOST=${COREOS_PRIVATE_IPV4} --volumes-from=deis-logger deis/controller
 
 [Install]
 WantedBy=multi-user.target