postgres-operator Spilo pods don't define liveness, readiness probes

...which leads to situations where e.g. if Patroni incessantly tries and fails to start postgres, this won't be visible in the pod's status (and hence in generic monitoring setups that don't specifically track e.g. the DB's replica count).

At the minimum I'd think the liveness probe should track the supervisord status (pod is live iff all services are running) and the readiness probe should track the Patroni state of the member (pod is ready iff the state is "running").

Jun 13 '19 16:06 multi-io

The absence of readiness probe impairs the PodDisruptionBudget logic, btw.

So I've defined those probes as follows. Maybe it's somewhat an overkill.

         livenessProbe:
          exec:
            command:
            - /bin/sh
            - -c
            - |-
              /usr/local/bin/supervisorctl status &&
              /usr/bin/pg_isready &&
              /usr/bin/curl -s localhost:8008 | /bin/grep '"state": "running"'
          failureThreshold: 8
          initialDelaySeconds: 60
          periodSeconds: 30
          successThreshold: 1
          timeoutSeconds: 5

        readinessProbe:
          exec:
            command:
            - /bin/sh
            - -c
            - |-
              /usr/local/bin/supervisorctl status &&
              /usr/bin/pg_isready &&
              /usr/bin/curl -s localhost:8008 | /bin/grep '"state": "running"'
          failureThreshold: 3
          initialDelaySeconds: 45
          periodSeconds: 10
          successThreshold: 1
          timeoutSeconds: 5