Re: [Bacula-users] How do I force Bacula to unlock
a drive? (SCSI Prevent/Allow Media Removal)

In the message dated: Tue, 23 Aug 2011 17:13:00 BST,
The pithy ruminations from Alan Brown on
<Re: [Bacula-users] How do I force Bacula to unlock a drive? (SCSI
Prevent/Allow Media Rem
oval)> were:
=> On 23/08/11 15:59, mark.bergman@xxxxxxxxxxxxxx wrote:
=> > The solution in our environment is to use the "sg_prevent" command
=> > (from the sg3_utils package) to release the SCSI Prevent/Allow Media
=> > Removal lock:
=> >
=> > sg_prevent --allow --verbose /dev/tape0
=> > sg_prevent --allow --verbose /dev/tape1
=> This works from a HBA other than the one which set the original lock???

I'm not quite sure what you're asking. As far as I know, the SCSI reservation
lock is specific to the combination of the SCSI device (a tape drive) and a
server, not to the HBA within the server.

The sg_prevent command must be issued from the server that originally
locked the drive. I don't know of a way to report on which server issued
the SCSI Prevent Media Removal command, so our usual method is to run the
"sge_prevent" command on each of the HA nodes that potentially locked the

Our servers have multiple fibre cards and paths to storage. For disk storage
devices with multiple fibre paths, a single server may use multiple HBAs to
access disk storage. However, each fibre tape drive is single-homed, so there's
only a single HBA per-server that connects to each tape drive. The tape
library has multiple fibre HBAs, but they are specific to tape drives--the
library (changer) and each drive are not multipath devices.

I'm eager to get information to clear up any of my misperceptions about SCSI



Get a FREE DOWNLOAD! and learn more about uberSVN rich system,

r administration capabilities and model configuration. Take
the hassle out of deploying and managing Subversion and the
tools developers use with it.
Bacula-users mailing list