-
Bug
-
Resolution: Not a Bug
-
Normal
-
ACM 2.10.Z
-
1
-
False
-
None
-
False
-
-
-
MCO Sprint 30, Observability Sprint 31
-
Moderate
-
None
Description of problem:
Observability pods should not inherit the defaultNodeSelector from the scheduler
Use case:
- oc get Scheduler cluster -oyaml | grep defaultNodeSelector
defaultNodeSelector: node-role.kubernetes.io/app= - oc get MultiClusterObservability observability -oyaml | grep -A2 nodeSelector:
nodeSelector:
node-role.kubernetes.io/infra: "" - The nodeSelector seen on the observability pods are:
node-role.kubernetes.io/app: ""
node-role.kubernetes.io/infra: ""
Version-Release number of selected component (if applicable):
How reproducible: Always
Steps to Reproduce:
- Set a defaultNodeSelector on the openshift scheduler:
oc get Scheduler cluster -oyaml | grep defaultNodeSelector
defaultNodeSelector: node-role.kubernetes.io/app=
- Set a different nodeselector on mco
- You will see that the observability pods inherit the defaultNodeSelector as well as from mco causing the observability pods to remain in pending state with the error:
X node(s) didn''t match Pod''s node affinity/selector
Actual results:
observability pods inherit the defaultNodeSelector
Expected results:
observability pods should not inherit the defaultNodeSelector by default
Additional info:
The only workaround is to declare the nodeselector on the open-cluster-management-observability namespace which will override the defaultNodeSelector setting.
- links to
-
RHSA-2024:135868 Red Hat Advanced Cluster Management 2.12.0 security and bug fixes