Description of problem

4.16 etcd v1 FlowSchema is unrecognized by the outgoing 4.15 Kube API server, causing noise during 4.15 to 4.16 updates.

Version-Release number of selected component (if applicable)

4.16.

How reproducible

Every time.

Steps to Reproduce

1. Install a 4.15 cluster.
2. Update to 4.16, while streaming CVO logs.
3. Search streamed CVO logs for Could not update flowschema "openshift-etcd-operator" ...the server does not recognize this resource, check extension API servers

Actual results

A bunch of hits, until the Kube API server finishes updating to 4.16.

Expected results

No distracting noise.

Additional info

Evgeni noticed this in OTA-1246, and Petr root-caused it.

OCPBUGS-22969 moved a bunch of FlowSchema to v1, and for most of them, it won't be a problem. But etcd and Kube API server update in parallel in run-level 20, so the CVO pushing out the etcd manifest can race Kube API server being new enough to understand it. One possible solution would be sticking with v1beta3 for 4.16 and moving to v1 in 4.17 (like our customers will have to do). Another solution would be moving the 4.16 manifest out to run-level 31 (e.g. by renaming 0000_20_etcd-operator_10_flowschema.yaml to 0000_21_etcd-operator_10_flowschema.yaml) after branch-fork, since the need would be unique to the 4.16 branch.