Merged into https://perconadev.atlassian.net/browse/PMM-2#icft=PMM-2.0 branch.

@Alexey Palazhchenko that's fine, so long as all services that should be in the RUNNING state under normal operations are checked

Please rely only on:

• response code is 200 = container is ready;

• any other response code or no code at all = container is not ready.

Do not really on other response codes, any response body (including empty JSON), etc. Supporting many failure modes requires a disproportionate amount of effort to the benefits.

@Roma Novikov we require v1/readyz to confirm that everything is ready, not just Prometheus. In particular, Grafana needs to be monitored as there is no clean way to check that we can interact with the API.

Also, you currently get an HTML 500 response by stopping pmm-managed, so adding that will presumably require NGINX adjustments to return a JSON 500 when requesting with Content-type: application/json

For @Roma Novikov to plan / prioritize future work.