Reservations on LUN.

[brian_31]

Team:

Have anyone faced the issue where a LUN was allocated from the storage (previously connected to a cluster) go into SCSI reservation and need to be released before being usable (on stand alone servers)

Thanks

Brian

[IT_2007]

Didn't understand your question well. What do you mean by SCSI reservation?

[nanan]

Hi brian
May I ask you to give more detail ??
It'd be batter if you explain what you wnat to do on your system....

[Serviceguard for Linux]

I haven't seen it, but it makes some sense.

Some clusters use SCSI reservation as a mechansim of protecting the cluster or protecting data. Depending on the cluster, there is some chance that this would "hang on" if you just shut down the servers without, at least, shutting down the cluster. You may even have to "delete" the cluster to get things into a clean state.

Because of the fact that some servers in a cluster will fail, the cluster SW would have some mechanism of knowing if it was OK the break the reservation.

[Mikko Niskanen_1]

At least in TruCluster environment, let's say that you have presented a MSA1000 array volume first to cluster, then want to re-use that volume without deleting it's contents to other Tru64 box, you'd need to run /usr/sbin/cleanPR script to clear out persistent SCSI reservation.

Otherwise, you'd run into following error:

scu> tur
SCSI Cmd = 'test unit ready', CDB length = 6, CDB bytes: 0 0 0 0 0 0
'test unit ready' failed, CAM status = 0x4 (CCB request completed with an error)
scu: 'test unit ready' failed on device 'MSA1000 VOLUME' at nexus [9/0/8]
SCSI Status = SCSI_STAT_RESERVATION_CONFLICT (0x18) - Target reservation conflict