-
Bug
-
Resolution: Unresolved
-
Critical
-
None
-
4.21
-
None
-
None
-
False
-
-
None
-
None
-
None
-
None
-
None
-
Proposed
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
Description of problem:
PROW CI - Hypershift pods are not coming up during hypershift-mce-install step
Version-Release number of selected component (if applicable):
OCP 4.21
How reproducible:
Always
Steps to Reproduce:
1.Trigger rehearsel for any job that consume hypershift-mce-install stage with 2.11 MCE version
Actual results:
+ oc get pods -n hypershift + grep 'operator.*Running' No resources found in hypershift namespace. No pods coming up in hypershift namespace.
Expected results:
+ oc get pods -n hypershift + grep -q 'operator.*Running' + _hypershiftReady=1 Pods coming up in hypershift namespace.
Additional info:
During CI execution, the pods in the hypershift namespace is not initialising when we try to test with 2.11 MCE and 4.21 OCP. On debugging we found that the hypershift-install-job-xxx from the open-cluster-management-agent-addon namespace is in error state. Reason: The two CRDs ipaddressclaims.ipam.cluster.x-k8s.io , ipaddresses.ipam.cluster.x-k8s.io getting generated have the version mismatch as spec.versions:- v1 & status.storedVersions:- v1beta2 . This causes Kubernetes to reject CRD reconciliation, which in turn prevents the Hypershift operator from coming up in the hypershift namespace. Workaround: The workaround which we did to delete these problematic CRDs and the pods & job in open-cluster-management-agent-addon , this forces the CRDs to be regenerated, and after regeneration the versions match correctly (v1beta1- v1beta2), allowing Hypershift to start.