Cluster wide pod rebuild from Kubernetes causes Trident's operator to become unusable
Applies to
- Trident 20.07.0, 20.10.0
- Openshift 4.5.13, 4.5.18
- Kubernetes 1.18
Issue
Since the pods are destroyed and created in a non-predetermined order, the orchestrator may find missing pods and attempt repair before the whole rebuild is finished.
The main error that can comes up during this is:
Warning FailedMount pod/trident-csi-<POD_ID> MountVolume.SetUp failed for volume "trident-csi-token-<POD_ID>" : secret "trident-csi-token-<POD_ID>" not found