Restarted TSM instance caused System State occupancy to double

[pjackson1]

Hi,

I was forced to halt and then restart a RHEL 7, 5.5.5.0 TSM instance on Monday AM during our main backup window due to the DB log approaching 100%. It is believed that most of our client on this instance were backing up at the time. The next day, Tuesday AM, we noticed that our TSM DB usage had increased from 222GB to 242GB, a 20GB increase or almost 10% (I realize this is over a the 150GB DB recommendation). Upon further analysis, we discovered that all 300 of our Windows 2008 clients (6.2.3.1) on this TSM instance had increased their system state file-count from about 70,000 files to 140,000. In other words, it looked like due to the TSM instance restart, we had "corrupted" the previous system state backup and all 300 of our Windows 2008 clients had performed a full system state backup. This hypothesis was further supported by the dsmsched log where we see an increase on average of 70k files backed up without any files assigned.

My question to the forum is has anyone else seen this behavior before? Why during the halt/restart of the TSM instance did the system state backup not pickup where left off?

Also, now that we increased our DB size due to the influx of new system state backups, we are going to try applying a custom management class to the system state backups with a 1-day expiration on old versions. Does this seem like the best approach to regain our space without completely deleting the system state filespace? IBM support hasn't come up with anything, yet.

Thanks!

 

[pjackson1]

Update

We still haven't figured out why TSM marks the previous system state as bad and sends a brand new system state. We have an open PMR for that with IBM.
The custom management class which we set up by adding the following to the client option set:

include.systemstate mcss

didn't seem to apply itself to the old system state backups. I valided this by running:

q ba "{Servername\SystemState\NULL\System State\SystemState}\"

and then seeing the Mgmt Class on the system state files listed as "DEFAULT".

Did we set up the mangement class correctly for the system state and if so, does it only apply itself to new backups? How can we apply it to old? Thanks.