High availability SCSI layout
From Linux NFS
It's currently possible to support simple active/passive high-availability NFS service. We'd like to make that's supported when using upcoming pNFS/SCSI support.
The plan for our first version is to treat migration as a reboot, as we do for other NFSv4 state: after migration, a client that attempts to use preexisting SCSI reservations or pNFS layout stateid's will get errors, those errors will trigger recovery behavior on the client which will cause it to reclaim any NFSv4 state and reacquire layouts as necessary.