ADSM-L

Re: Two questions on 5.1.x --> 5.3.x

2005-10-05 11:28:27
Subject: Re: Two questions on 5.1.x --> 5.3.x
From: Alexander Lazarevich <alazarev AT ITG.UIUC DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Wed, 5 Oct 2005 10:28:09 -0500
On Wed, 5 Oct 2005, Paul van Dongen wrote:

Hello all,

   Sorry to bother you not with one, but three questions...

   I am in the process of upgrading a customer's TSM Server from 5.1.6.4 to 
5.3.x. I am taking some precautonary measures due to the size of the DB 
(240GB). Here is what I thought of:

1) Backup DB and apply 5.1.10 fixpack (Scheduled for today)
2) Run CLEANUP BACKUPGROUPS
3) Upgrade HP-UX (Now 11.0) to 11i and atdd
4) Backup DB and install 5.3.0.0
5) Install 5.3.2

So question one is: Is this path correct or should I think of installing 5.2.x 
in between? And would you recommend 5.3.2 or a more conservative 5.3.1.6?

I went from 5.1.9 to 5.3.1 on win 2K server, with no problems. I did not
migrate to 5.2 first. I just followed all the v5.3 README instructions
carefully, as well as the instructions in this cool audio power point
presentation:

http://www-1.ibm.com/support/docview.wss?rs=663&context=SSGSG7&q1=tsm+5.3+upgrade&uid=swg21214007&loc=en_US&cs=utf-8&lang=en

Make sure to backup the volumehist and dev stuff too, but all that is in
the README and the presentation.

Good luck.

Alex





For question 3, the following info:

   The database was behaving VERY badly. It had been allocated as 240 (!) 
volumes 1GB each on 3 separate filesystems (EMC disks). I already changed it to 
8 33GB volumes on 8 filesystems. Performance became a little better (backup DB 
went from almost 4 hours to 2h50m), but it still suffers bad on expirations. It 
takes 3+ hours to expire a million objects, and by analysing actlog I found 
that it takes almost a month to expire twice the same node/filespace. At these 
moments, sar -d shows one of the DB volumes at nearly 100% util, but with a 
rather low number of operations (long seeks? DB fragmentation too bad?).

   So question is: Should I leave the DB this way, or use the other disks given 
to me by the storage admin: 32 8GB slices on separate disks. Take this slices 
and make 8 33GB striped LVs (4 slices each), create the filesystems and put the 
DB volumes there?

Thanks a lot for your attention,

Paul van Dongen


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