Average Time to Restore a TSM Server

[STSM]

Hello everyone, I have three question, just general please no horror stories. I am planning a DR Drill and need some answers.

Now I know this question really is based on the size of the database and the hardware.

Just looking for a general answers.

What is the Average time to rebuild TSM from nothing? (no OS, and TSM DBBackup is on tape)

How many times a year do you test the rebuild of TSM?

How many people from the backup team is needed to rebuild the server(s)?

Does it make sense to have TSM installed somewhere else and just restore the DB when needed? If so what is the average time?

 

[Cyclone]

In all honesty, the answers are relative to each system. There are small TSM systems, and there are large TSM systems. So the time it takes is relative.

But this is why you test. To answer all of those questions. Plus as you test more often you will edit your procedures and streamline them to go faster.

Average time to build a TSM server from scratch. I would say about 6 - 10 hours depending on how fluid you are. The restore of the db can take a while if it is large.

How many times a year should you test? About twice a year os good.

How many people from the back up team are needed? Answer: relative. Again, it could be just one person who is able to install the OS, TSM, configure it, run the drm, restore the db, etc. Or you may have to split up those duties upon the talents of your workers.


Does it make sense to have a hot spare ready to go? Depends on what your acceptable "down time" is. If it doesn't matter that your TSM server is unavailable for 24 hrs, then you don't need a hot spare. But if you need to have those restores running in a matter of a few hours, then it may be a good idea.

 

[STSM]

Thanks for the feed back. I will take other suggestion and post so please keep them coming.

 

[evilution]

You should be able to answer the question about hardware and OS config.

As for TSM its simply clicking on the EXE and next a few times. Then running a command to restore the DB. I'd say the DB comes back at a rate of 20GB per hour. But once its up there are audits to run and those can take hours depending on what kind of clean up is required.

 

[ad283m]

You should be able to answer the question about hardware and OS config.

As for TSM its simply clicking on the EXE and next a few times. Then running a command to restore the DB. I'd say the DB comes back at a rate of 20GB per hour. But once its up there are audits to run and those can take hours depending on what kind of clean up is required.

He is absolutely right.

For example, a 70 G database the restore was like 2 or 3 h but the audit took near 12 h

So it depends on how the database was and the h/w, once we used LTO1 and 70 GB the database and it was painful a day or so, compared with a 250 GB and lto3 and it was like 4 - 5 hours to bring the server back online

 

[ad283m]

I mean, if the backup is nice and there are no problems, then no database audit, if there is a message, then an audit must be done, aldso how long between the last backup and the crash have passed will determine how hard is to recover (you know the post steps after the restore db)

 

[STSM]

Hello all, I understand that TSM can take some time restoring the Database, I also understand that if there is a problem with the database it will restore even with that problem still there, and the only way to truely know if the database has problems is to dump then load, at the end of the load if issues are found TSM will tell you to audit the database. Seems to me that is a real problem? Am I right about that?

 

[ad283m]

Ok STSM in my experience, usually the database does not need to be audited.

Regarding the dump-load, that's pretty much the last resource, so it's like the 5% of the cases, in which things are in a pretty bad shape.

This are the post steps after you restore the database:
https://www-304.ibm.com/support/docview.wss?uid=swg21294024

So there is an audit of the volumes.

I repeat, the database usually does not need to be audited, unless it asks you to do it. And it's not normal procedure.

If the guys here have any additional input would be greatly appreciated.