Bug: restic-wait mounts shared volumes in only one statefulset member

[davidcmitchell]

Contact Details

david.mitchell@ibm.com

Describe bug

When restoring a statefulset backed by managed-nfs-storage, the restic-wait init container mounted a shared volume in only one member (meta is the shared pvc):
c-db2wh-test-db2u-0

  restic-wait:
    Container ID:  cri-o://57c509cebc5296859d892e3d20f25f79952fe7638305776bfae0c5821b8f8ab4
    Image:         quay.io/konveyor/velero-restic-restore-helper@sha256:c16248d027c09cd5b7360ae38484df88ca1cc01f9e470e3eade95076ee1a6b84
    Image ID:      quay.io/konveyor/velero-restic-restore-helper@sha256:55b0bc00b8603f8ac55fee791451a089dbadc648af498dc76647cf280b711ea2
...
    Mounts:
      /restores/data from data (rw)
      /restores/meta from meta (rw)
      /restores/tempts from tempts (rw)
      /var/run/secrets/kubernetes.io/serviceaccount from kube-api-access-hmj5t (ro)

c-db2wh-test-db2u-1

    Mounts:
      /restores/data from data (rw)
      /restores/tempts from tempts (rw)
      /var/run/secrets/kubernetes.io/serviceaccount from kube-api-access-hmj5t (ro)

As a result, one of the pods became Ready before this volume was fully restored, causing problems.

Here is the volumes defn in the statefulset:

        volumeMounts:
        - mountPath: /mnt/blumeta0
          name: meta
        ...
      volumes:
      - name: meta
        persistentVolumeClaim:
          claimName: c-db2u-test-meta

What happened?

It is unclear if this has ALWAYS been the behaviour of the restic-wait init-container for statefulsets or this is a new issue. For instance, how can a member of a statefulset proceed before a shared volume is restored?

OADP Version

0.5.x (Stable)

OpenShift Version

4.10

Velero pod logs

No response

Restic pod logs

No response

Operator pod logs

No response

New issue

• This issue is new

[Kaustubh-Sabnis]

Would it be possible to share any ETA when this can be fixed ?

[weshayutin]

https://issues.redhat.com/browse/OADP-1489
Closing github issue.

[davidcmitchell]

@weshayutin What's the resolution for this issue? That issue you linked doesn't look related.

[davidcmitchell]

@weshayutin Can we please reopen?