On Aug 21, 2008, at 21:44 , Nick Laflamme wrote:

How common is it for sites to cluster TSM servers?

I know of some organizations that at one time or the other used to run
TSM in a HACMP cluster. Most (if not all) decided that the current IBM
p series hardware has so much built-in resilience (or redundancy) that
hacmp only adds complexity without much benefits, esp in multi-CEC
p570 systems.

So yes, there used to be quite a few, and currently, well fewer and
fewer. Hardware failure in a p5 system is so rarely fully service
affecting that it's just not worth it.

Now, remember it's only backups. You could possibly argue having a
cold-standby TSM server on a remote location, running some automated
db restore script that will allow you to get started again in very
little time with a very short rpo, but HACMP or other fail-over
clusters make no sense to me.

I'm looking for my next TSM position and have found at least one
situation, perhaps two, where the would-be client wants TSM
administrators who are familiar with CSM in an AIX/p-series
environment. As far as I can tell, that implies setting up the TSM
servers to fail-over so that if one fails, a hot-spare immediately
covers for it. I've never heard of this, and I've worked in a couple
of large, multi-TSM server environments.

My working theory is that someone took an AIX administrator set of
requirements as a template and didn't remove enough extraneous
buzzwords, but if more sites than I realize are clustering TSM
servers, I guess I should know that, too.

Thanks,
Nick

--
Met vriendelijke groeten,

Remco Post
remco AT pipsworld DOT nl