Troubleshooting
Problem
A Tivoli Storage Manager server running on Sun Solaris 10 and configured with a DISK storage pool on a ZFS file system suffers from poor performance when multiple clients back up to the ZFS DISK storage pool.
Cause
Sun Solaris Bug ID 6596237
Diagnosing The Problem
Collect Tivoli Storage Manager server performance data using instructions in the following document:
http://www-1.ibm.com/support/docview.wss?uid=swg21260310
The above document contains a server.pl perl script which, when run, collects server instrumentation data. This data can be analyzed to determine performance bottlenecks. In this situation, the collected server instrumentation data will show poor performance in "Disk Commit". For example :
Thread 25 DiskServerThread parent=0 (Sun TID 36) 11:06:02.693-->11:26:08.172
/myzfs/tsmpool/diskvol1
Operation Count Tottime Avgtime Mintime Maxtime InstTput Total KB
------------------------------------------------------------------------
Disk Write 261 0.733 0.003 0.000 0.102 80051.7 58708
Disk Commit 257 204.026 0.794 0.042 2.537
Thread Wait 256 1000.650 3.909 0.000 148.974
Unknown 0.067
------------------------------------------------------------------------
Total 1205.478 48.7 58708
The above data shows that "Disk Commit" was high for volume /myzfs/tsmpool/diskvol1.
Resolving The Problem
The poor performance in "Disk Commit" is a result of Sun Solaris Bug ID 6596237 (Stop looking and start ganging). It has been reported that Solaris Bug ID 6596237 is addressed in Solaris 10 10/09 (Update 8) or in kernel patch 141444-09 for SPARC and kernel patch 141445-09 for x86/x64. Consult Solaris support for further information on this defect.
Was this topic helpful?
Document Information
Modified date:
17 June 2018
UID
swg21390367