There are several problems on SRIOV pods after doing a system reboot on a SNO cluster.

SRIOV pods appear as duplicated. The problem seems to be that SRIOV pods that exist before the reboot are not deleted properly.

oc get pods:
...
test32-deployment-6b5c896c96-6r9kf     6/6     Running                    0          19m
test32-deployment-6b5c896c96-m5qzv     0/6     UnexpectedAdmissionError   0          32m
...

New bug seen in 4.12.20. It hasn't been seen before.

100% of times after doing system reboot, graceful shutdown or via redfish.

1. Deploy DU application with several SRIOV pods
2. Reboot the system (https://docs.openshift.com/container-platform/4.12/backup_and_restore/graceful-cluster-shutdown.html) or reboot via redfish
3. Check SRIOV pods after reboot

Some SRIOV pods appear as duplicated and with errors:

oc get pods:
...
test32-deployment-6b5c896c96-6r9kf     6/6     Running                    0          19m
test32-deployment-6b5c896c96-m5qzv     0/6     UnexpectedAdmissionError   0          32m
...

oc get pods --> pods running normally

System impact: Manual cleanup of SRIOV pods of the DU is required after any kind of reboot.


Old pod32 describe:

oc describe pod/test32-deployment-6b5c896c96-m5qzv

  Warning  UnexpectedAdmissionError  25m   kubelet            Allocate failed due to no healthy devices present; cannot allocate unhealthy devices openshift.io/pci_sriov_net_llscu, which is unexpected