Would either of these be of any help?
MEDIA_UNMOUNT_DELAY
To specify a Media unmount delay property indicates that the unloading of media
is delayed after the requested operation is complete. Media unmount delay
applies
only to user operations, to include backups and restores of database agent
clients,
such as those running NetBackup for Oracle. The delay reduces unnecessary media
unmounts and the positioning of media in cases where the media is requested
again
a short time later.
The delay can range from 0 seconds to 1800 seconds. The default is 180 seconds.
If
you specify 0, the media unmount occurs immediately upon completion of the
requested operation. Values greater than 1800 are set to 1800.
RB_MPX_GROUP_UNLOAD_DELAY
The RB_MPX_GROUP_UNLOAD_DELAY parameter indicates the number of
seconds that nbrb waits for a new job to appear before a tape is unloaded.
(Default: 10 seconds.)
RB_MPX_GROUP_UNLOAD_DELAY=10
This setting can help avoid unnecessary reloading of tapes and applies to
all backup jobs. During user backups, nbrb uses the maximum value of
RB_MPX_GROUP_UNLOAD_DELAY and the Media mount timeout host
property setting when nbrb unmounts the tape.
During restores, Media mount timeout is used, not
RB_MPX_GROUP_UNLOAD_DELAY.
Many regards,
Pat Whelan
Principal Remote Product Specialist - NBU
Business Critical Services
Symantec Corporation
www.symantec.com
Mobile: +44(0) 771 492 3085 (Preferred)
Office: +44(0) 118 943 6715
Email: patrick_whelan AT symantec DOT com
-----Original Message-----
From: veritas-bu-bounces AT mailman.eng.auburn DOT edu
[mailto:veritas-bu-bounces AT mailman.eng.auburn DOT edu] On Behalf Of David
Stanaway
Sent: 05 December 2013 10:28
To: veritas-bu AT mailman.eng.auburn DOT edu
Subject: [Veritas-bu] Queued duplication jobs churning - tape mounts increasing
We had some recent hardware issues with tape robot and had to re-do robotics
config on NBU.
Meanwhile we had storage unit marked as inactive for the SLPs using that robot
for a duplication stage.
Now that the storage unit has been marked as active we have about 300 queued
duplication jobs and it looks like after each image copy in the active
duplication job finishes the job gets superseded by another backup job which
unmounts the tape and remounts it then does 1 image.
This is making duplication performance really bad and the tape mounts for the
active media are going through the roof. They go from 0 to about
400 before media fills up.
We didn't have this issue before the robotics reconfig I don't think.
Scratching my head on this one. It may go back to a change in behavior after an
upgrade from 7.1.0.3 to 7.5.0.4 also
Info on the storage unit config.
Dell ML6010 with 2 LTO4 drives presented to 2 media servers
There is a 2 host SUG with the 2 drives.
The target media group has max non-full media set to 2
At any time, I am seeing 4 active duplication jobs oscillating with no drive
available with normally only 1 active at any given time due to unmount/remount
overhead. Obviously only 2 can actually be writing to media at any given time.
I tried taking one of the hosts out of the SUG, but that didn't help, still
have 4 active jobs juking it out for the 2 drives and thrashing the robot and
media mounts.
Anyone have a suggestion? This is killing the robot and lifecycle policy
windows.
_______________________________________________
Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
_______________________________________________
Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
|