TSM recovery with multiple Instances

[janichols6]

Ideas needed for DR testing.
We do not have an offsite facility or a proper test server, but the powers that be want to know what it will take to recover our TSM server.

Here is the dilemma I ran into already, we are running on AIX, TSM5.5 with 5 instances.

I got another AIX LPAR with a different IP address, loaded the OS of my existing TSM server and zoned the existing tape library to the test lpar.

I had to do a crossdefine to the Production Library Manager so I could use the Tape Drives.

I did a database restore for the first instance on the test lpar which was successful, but when I shut it down it also shutdown the Production Instance. I'm thinking the crossdefine did me in, but it's the only way to get to the drives.

Can you rename the instances in the devconfig and dsmserv files before you do the restore? Or will I need to have the Production instances down while I recover on the Test lpar, then rename them and then bring the Production instances back online? - that will be a big pain.

 

[moon-buddy]

Unfortunately, you cannot rename any TSM server as the database restore will be looking for the right TSM server name.

This is why in a DR test, you need to restore on an isolated environment; or if this isn't possible, shutdown the Production TSM server/server instance.

 

[janichols6]

Ya, that's what I thought:redface:

Which is what I told the PTB

 

[Eldoraan]

Be careful on restored instance if you're using the production library & media! If you have an instance running as a library manager, and you bring up a restored library client, it may be out of sync with current library inventory. You could have the restored instance scratch a tape it thinks is still pending scratch, only to find out that tape has been reused by the instance between DR's db backup and present. You can end up overwriting good data when tapes in Pending state on the DR instance are released as scratch when it starts up. The library manager instance will be more than happy to return the volumes to scratch even though they have been reused after they went scratch in production.