Disk snapshots with XFS filesystem cannot be used

[jsafrane]

It is not possible to restore a snapshot of a XFS volume and use it on the same node as the original volume.

Steps to reproduce:

1. Create a StorageClass with csi.storage.k8s.io/fstype: xfs.
2. Create PVC A + Pod A using it, store some data on the provisioned volume.
3. Stop the Pod A
4. Take a snapshot of the PVC A, restore it into a new volume as PVC B.
5. Run both Pod A (with PVC A) and Pod B (with PVC B) on the same node.

Actual result: one of the pods can't be started:

MountVolume.MountDevice failed for volume "pvc-25a7129e-2cd7-4170-9050-550057ed0e20" : rpc error: code = Internal desc = mount failed: exit status 32
Mounting command: mount
Mounting arguments: -t xfs -o shared,defaults /dev/vde /var/lib/kubelet/plugins/kubernetes.io/csi/pv/pvc-25a7129e-2cd7-4170-9050-550057ed0e20/globalmount
Output: mount: /var/lib/kubelet/plugins/kubernetes.io/csi/pv/pvc-25a7129e-2cd7-4170-9050-550057ed0e20/globalmount: wrong fs type, bad option, bad superblock on /dev/vde, missing codepage or helper program, or other error.

Kernel says:

[17567.886004] XFS (vde): Filesystem has duplicate UUID b5820d11-d42d-4fc1-8e4a-ae431b33039d - can't mount

Expected result: both pods can run.

See container-storage-interface/spec#482 for details.

There is a simple fix, add nouuid mount option to all XFS mounts to ignore the UUID. Here is how we fixed it in AWS EBS CSI driver: https://github.com/kubernetes-sigs/aws-ebs-csi-driver/pull/913/files#diff-4f4d9c2e0ec5c4d3b4ac2d79abd70ee21bacff65ba05c059396921996dbf6607R673

[mowangdk]

ok, i see, i will fix it ASAP