Currently, our /v1/readyz readiness pmm-managed API checks only Prometheus status (and, indirectly, returns nothing if nginx, pmm-managed, or PostgreSQL is down). Managed services require a check for Grafana too.

DoD

• /v1/readyz returns an error if Grafana is no ready (down, starting up, or shutting down).

Implementation

• Check what Grafana Health API returns when Grafana is starting up or shutting down.
• Add a method to our Grafana client to access that API. We might need to expect a response body for that, not only the status code.
• Use that method in readiness API.

Discussion

• We are not checking `supervisorctl status` output (as used by update mechanism) as this is too brittle and a constant source of various tricky update bugs.