[Question] what's expected behavior when nfs-server pod is restarted?

[ileixe]

It's not a bug rather a question. We're looking for stable solution for RWX and trying to use this project with ceph-block.

We wonder how's the failover time for pod when killing nfs-server pod on purpose. Surprisingly, the pod never come back work (hang on the shell) although nfs-server is restarted normally.

What's the expected behavior in this case? Is it somehow related to ceph-block (if I choose other block backend, nfs-provisioner can fallback well?)

[kstreee-furiosa]

Tried to delete pod, but got following event

Events:
  Type     Reason         Age                  From     Message
  ----     ------         ----                 ----     -------
  Warning  FailedKillPod  2m (x1001 over 12h)  kubelet  error killing pod: failed to "KillContainer" for "workspace" with KillContainerError: "rpc error: code = Unknown desc = Error response from daemon: cannot stop container: ad747952b359edfd96c1972c3ada8f21754a90c7a68592d0102718754ffb95b3: tried to kill container, but did not receive an exit event"

[dsharma-dc]

@ileixe Not neccesary recommendation but have you tried using any other backend like localpv-hostpath or openebs-jiva etc.
Do you have any logs you can share to take a look?

[ileixe]

We are guessing it's from lower level implementation(kubelet) not provisioner itself: kubernetes/kubernetes#31272