> I did not check what pods run on the remaining node and how much CPU is left. The scheduler calculations could be right, but then there is a serial tests that leaves the cluster in a state when there is a node that does not have 50m CPU available. This is IMO a bug and it needs to be fixed. which test? is this being pursued?
Search "EmptyDir wrapper volumes should not cause race condition when used for configmaps" string in" - https://testgrid.k8s.io/redhat-openshift-ocp-release-4.5-informing#release-openshift-ocp-installer-e2e-azure-serial-4.5 looks good (rarely red) - https://testgrid.k8s.io/redhat-openshift-ocp-release-4.4-informing#release-openshift-ocp-installer-e2e-azure-serial-4.4 looks good (rarely red) - https://testgrid.k8s.io/redhat-openshift-ocp-release-4.3-informing#release-openshift-ocp-installer-e2e-azure-serial-4.3 looks good (rarely red) - https://testgrid.k8s.io/redhat-openshift-ocp-release-4.2-informing#release-openshift-ocp-installer-e2e-azure-serial-4.2 red more often but still not very frequently
Given the frequency of flakes reduced to almost zero, I am closing the issue.