define readinessProbe on statefulSet

[FxKu]

With #1760 merged, we should add a readinessProbe to the statefultSet because the ConfigMaps option defines a selector on the master service. Label selectors incl. readiness probes. We will add it either way.

#1760 also introduced a new suffix for deleting Patroni cluster objects which will result in an error for the new "leader" suffix if endpoints are used. Instead, the operator should just l log a warning and try to delete as much as it can. Another bug was introduced by not deleting all services if endpoints are used.

Other changes:

• user variables when specifying container ports
• check configmaps option when syncing ClusterStatus not omit endpoints
• remove endpoints from RBAC (provide RBAC manifest for OpenShift)

[jopadi]

👍

[FxKu]

👍

[beneso]

Hi, thanks for the commit!
I have a question: are you sure of the port? Container is never ready with this probe, HTTP 503. While if we try the same path with port 8080, we get HTTP 200.

root@acid-grafana-postgres-0:/home/postgres# curl -vvvv $POD_IP:8008/readiness
*   Trying 10.0.2.66...
* TCP_NODELAY set
* Connected to 10.0.2.66 (10.0.2.66) port 8008 (#0)
> GET /readiness HTTP/1.1
> Host: 10.0.2.66:8008
> User-Agent: curl/7.58.0
> Accept: */*
> 
* HTTP 1.0, assume close after body
< HTTP/1.0 503 Service Unavailable
< Server: BaseHTTP/0.6 Python/3.6.7
< Date: Tue, 31 May 2022 15:43:49 GMT
< Content-Type: application/json
< 
* Closing connection 0
<json>

vs port 8080:

root@acid-grafana-postgres-0:/home/postgres# curl -vvvv $POD_IP:8080/readiness
*   Trying 10.0.2.66...
* TCP_NODELAY set
* Connected to 10.0.2.66 (10.0.2.66) port 8080 (#0)
> GET /readiness HTTP/1.1
> Host: 10.0.2.66:8080
> User-Agent: curl/7.58.0
> Accept: */*
> 
< HTTP/1.1 200 OK
< Content-Type: application/json
< Date: Tue, 31 May 2022 15:43:44 GMT
< Content-Length: 3375
<json>

Thanks!