restore db after disk crash

[wesd]

Hi,

I'm trying to write a DR plan for my 5.4.1 tsm server. The user guide talks about point-in-time recovery and mentions that, if the original database or logs are lost, you need to run dsmserv format. Since my database and logs are on different disks, I need to document the situation where the database disks have been lost but I have a full backup (run daily) plus a complete, undamaged set of recovery logs.

Once I create a new set of volumes for the database, do I need to run anything other than dsmserv restore db? The point-in-time document suggests that a dsmserv format is necessary (loss of database OR logs) but won't that destroy the logs that I need for the restore?

 

[chad_small]

dsmserv format is necessary since it will create the dsmserv.dsk file which is required for TSM to know where the DB and Log volumes are located on startup. Just formating them yourself with not do it.

 

[wesd]

But does that not mean that the logs will be destroyed by the formatting? This would suggest that placing the database and logs on different disks is pointless since losing the database will mean destroying the logs anyway.

 

[travelsick]

sounds like you wouldn't be a dsmserv format for the log volumes, just the database volumes.

Don't you already have a backup copy of your dsmserv.dsk?

 

[wesd]

I think I may be confused about what dsmserv format actually does. I thought it formatted the disks which hold the database and logs but, instead, does it simply create a dsmserv.dsk file? Maybe it's the word 'format' that's confusing me.

In the DR circumstances that I'm trying to document I will have -

A copy of the dsmserv.dsk file before the crash
A full set of intact log disks
No database disks

If I use dsmfmt to recreate the disks exactly as they appear in the dsmserv.dsk file, is this all I need to be able to do a complete data restore with the database coming from the last full tape backup and rolled forward using the log files?

If I use dsmfmt to create the disks differently eg. the original dsmserv.dsk has 10 25GB disks but I create single 250GB disk (all my disks are san luns) what do I need to do?

I know this sounds like I'm trying to make things more difficult than they need to be but this type of disk loss if what I've been told to document. In practice, all my disks are mirrored to another datacentre and I can't think of a situation where I would lose both mirrors but, nevertheless, this is what I have to document.