Ceph RBD: Restore the filesystems UUID on the volume

[roosterfish]

Ceph RBD snapshots are a read-only logical copy of the parent volume. Rewriting the filesystems UUID on the snapshot itself fails.
That was observed in #12743

Fixes #12744

[tomponline]

This looks good and I'll merge this as its a vast improvement on whats there now.
However if you take a look at the LVM driver for this same functionality, it moves the target volume to a temporary name before restoring the snapshot to the target volume name so that if any of the restore or uuid regeneration steps fail it can delete the partially restored volume, swap the old target volume back and not leave the target volume in an corrupted state.

Is this something we can do with the ceph driver too?

[roosterfish]

This looks good and I'll merge this as its a vast improvement on whats there now. However if you take a look at the LVM driver for this same functionality, it moves the target volume to a temporary name before restoring the snapshot to the target volume name so that if any of the restore or uuid regeneration steps fail it can delete the partially restored volume, swap the old target volume back and not leave the target volume in an corrupted state.

Is this something we can do with the ceph driver too?

Thanks, I'll add a note. Especially the reverter pattern can be used more frequently. What I also found is that the VM's filesystem volume doesn't get restored. Looks to be missing also for other drivers if I am not mistaken.