Error "Command failed, re-intializing driver" during device interaction or backups (4047286)

During a device interaction or backup to a tape device connected to a Windows machine, the log file will show a message like:
Error 2010/10/21 13:45:48 0 Media Command failed, re-initializing driver

with an additional log context of:
SCSI :1936 14 0 124542 CMD REZERO/REWIND [ 01 00 00 00 00 00 ]
CAMNT :1936 38 1 124548 (NT) Device I/O control failed (0,1117)
CAMNT :1936 40 1 124548 (NT) Windows NT Kernel SCSI Passthrough unrecoverable NT kernel error
CAMNT :1936 41 1 124548 (NT) SPT 000012fa14: 2c 00 00 00 05 00 06 18 02 00 00 00 00 00 00 00
CAMNT :1936 41 1 124548 (NT) SPT 000012fa24: 84 03 00 00 00 00 00 00 30 00 00 00 01 00 00 00
CAMNT :1936 41 1 124548 (NT) SPT 000012fa34: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
CAMNT :1936 42 1 124548 (NT) SCMD 00008a34cc: 01 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
SCSI :1936 15 0 124548 CMD failed: errno 1117

This is the response to a positioning command, but similar errors could be seen for any operation, taking drives/libraries offline. Sometimes, this can occur for writes in the data transfer stage as well.

After these errors, subsequent move commands might fail with
CHGSCSI :2744 22 0 114335 CMD MOVE MEDIUM [ a5 00 00 00 01 e1 00 2a 00 00 00 00 ]
CHGSCSI :2744 24 0 114404 CMD STATUS 'CHECK CONDITION' [2]
CHGSCSI :2744 25 0 114404 CMD SENSE KEY 'ILLEGAL REQUEST' [5] - ASC 53 ASCQ 02
CHGSCSI :2744 74 0 114404 CMD SENSE INFO 'MEDIUM REMOVAL PREVENTED'

This can have been caused by the library being in an illegal state after the earlier errors on the bus, preventing future correct behavior of the drives/library and preventing "check" commands from succeeding.

A command or data message does not get acknowledged on the SCSI bus in time, and instead the operating system decided to reset the bus. The most common reason is hardware-based, like cabling, termination etc, but it could also be HBA firmware, driver revision or in the case of a SAN zoning and/or switching on Fiber Channel configurations.
 

Occasionally, this can be recovered from by stopping/starting the service and/or rebooting library and server, but depending on the the underlying error, it might reoccur. Ultimately, it needs to be addressed by HBA and operating system vendor.

BakBone SFDC Solution Number: 00001556