SCSI reservation conflicts reported in environment. Drives going into "DOWN" state.

Problem

Drives going into a "DOWN" state when running backups.

Issues such as the following are reported in the messages log:

Aug 28 10:48:35 server_name kernel: st 8:0:3:2: reservation conflict

Aug 28 10:29:55 server_name kernel: st 8:0:1:2: reservation conflict

Error Message

Issues such as the following are reported in the messages log:

Aug 28 10:48:35 server_name kernel: st 8:0:3:2: reservation conflict

Aug 28 10:29:55 server_name kernel: st 8:0:1:2: reservation conflict

Cause

Issue reported with tape drives going into a DOWN state when backups were run. Examination of messages log on NetBackup hosts showed SCSI reservation conflict errors.

vmoprcmd -crawlreleasebyname did not clear the reservations so something external to NetBackup had reservation of the drives.

Solution

NPIV (essentially shared fibre connections - multiple FC initiators sharing a single physical port) can cause issues in SSO (Shared Storage Environments) if the underlying hardware is not inter-compatible. It is important to confirm with hardware vendors that devices in the environment are inter-compatible and drivers/firmware are up to date and support NPIV. In SSO environments it is imperative that SCSI traffic passes through the network/SAN infrastructure unmolested and without corruption, otherwise issues such as SCSI reservations can arise due to devices not releasing their reservations correctly and/or being unable to acquire them.

Applies To

Solaris 10

NetBackup 7.1.0.3

Cisco UCS B230 M2 Blades

NPIV (N_Port ID Virtualization - see http://en.wikipedia.org/wiki/NPIV)