Description of problem:
According to https://kubernetes.io/docs/concepts/workloads/pods/disruptions/#pod-disruption-conditions, there are total 5 reasons for DisruptionTarget type: PreemptionByScheduler, DeletionByTaintManager, EvictionByEvictionAPI, DeletionByPodGC, TerminationByKubelet. However, we only fixed two:
For bug https://issues.redhat.com/browse/OCPBUGS-45490, we fixed `TerminationByKubelet` reason.
For bug https://issues.redhat.com/browse/OCPBUGS-42150, we fixed `DeletionByTaintManager` reason.
So, we still need to fix others:
PreemptionByScheduler EvictionByEvictionAPI DeletionByPodGC
Discussion on Slack: https://redhat-internal.slack.com/archives/C06KP34REFJ/p1733709111219019?thread_ts=1733452156.229049&cid=C06KP34REFJ
Version-Release number of selected component (if applicable):
OCP 4.14 +
How reproducible:
always
Steps to Reproduce:
1. 2. 3.
Actual results:
CatalogSource cannot be recreated.
Expected results:
CatalogSource can be recreated.
Additional info:
- relates to
-
OCPBUGS-45490 Evicted Pods owned by Catalogsource are not rescheduled
- Verified
-
OCPBUGS-42150 [4.14]OLM catalogsource pods do not recover from node failure when registryPoll is none
- Closed