This is a bug affecting Quay operator 3.12.2.

Quay operator never stops the reconciling process when there are multiple ingress controllers defined in the OCP cluster.

This leads the pods to be unstable since new replica sets are continuously created.

From Quay operator logs, we observed that the operator continuously detects a change to the cluster hostname:

{"level":"info","ts":"2024-11-22T09:31:22Z","logger":"controllers.QuayRegistry","msg":"Detected cluster hostname apps.example.com"} 

{"level":"info","ts":"2024-11-22T09:32:23Z","logger":"controllers.QuayRegistry","msg":"Detected cluster hostname intern-apps.example.com"}  

By taking a look at the source code, it appears that the operator picks the first ingress in the list [1]
However, the order of ingresses is not guaranteed to always be the same.
Consequently, the operator might pick a different ingress and trigger the reconciliation process.

The issue does not happen when there is only one ingress controller defined.

The workaround is to set route and tls components as unmanaged and create a custom route manually.

[1] https://github.com/quay/quay-operator/blob/25bc2dc1ae8eed37f41335bd3e1b08efc1e4540a/controllers/quay/features.go#L268-L271