ADSM-L

Re: Archive problems

2001-09-06 11:21:57
Subject: Re: Archive problems
From: "Remeta, Mark" <MRemeta AT SELIGMANDATA DOT COM>
Date: Thu, 6 Sep 2001 11:22:36 -0400
Martha, here is a response I received from Tivoli support:

Hi Mark,
       I know I haven't spoken to you in a while regarding this problem,
and I apologize for that; but I have been able to determine the procedure
to help you improve your archive performance. The 3.7.2 and 4.1.2 or higher
TSM clients are designed to no longer backup duplicate directories.  That
is, directories with the same name in the same path with the same
description will only have 1 copy on the TSM server.  What we need to do to
ensure all old duplicates are deleted and the new, more efficient archive
tables are being used by your nodes, is run the following commands from a
TSM Admin command line:

CLEAN ARCHDIR <nodename> SHOWSTATS
DELETE ARCHDESCRIPTIONS <nodename>
CONVERT ARCHIVE <nodename> REORG=YES

     Please note that the DELETE ARCHDESCRIPTIONS command DOES NOT remove
all the descriptions from your archive packages, it only deletes entries
from some of the descriptions tables.  Once you run CONVERT ARCHIVE, the
description entries are rebuilt from the file data using the new, more
efficient description table, so you will still be able to access all your
archives by description.  When this is completed, you can run CLEAN ARCHDIR
<nodename> SHOWSTATS again to view changes.  SHOWSTATS output may still
show duplicate directories, but these are not true duplicates because they
have different descriptions, so the output is correct.

     After all old nodes have been converted using this process, you should
see improved performance.  New nodes using 4.1.2 and higher clients
connecting to 3.7.4 and higher or 4.1.2 and higher TSM servers will already
be using the new tables and do not need to go through the above conversion
process.  Even after this conversion, as the number of archives continues
to grow, you may still experience slowdown as it becomes more and more time
consuming to query and insert new archives into the existing tables.  As
you have already seen, using a new nodename for archives does workaround
this problem.


<Prev in Thread] Current Thread [Next in Thread>