Details
-
Story
-
Resolution: Unresolved
-
Normal
-
None
-
None
-
False
-
-
False
Description
The Controller Runtime framework has functionality to run a Reconcile event re-queue even if all watched objects are detected as healthy/reconciled. This could give us extra assurance that we are in a happy state, so let's investigate adding a lenient/occasional re-queue.
When implementing we should also check/validate performance concerns, as by-design we would be running and order of magnitude more reconcile events (and therefore kube API calls)
Migrated from GitHub: https://github.com/opendatahub-io/data-science-pipelines-operator/issues/202