TSM Server Crashed

[wgadmin]

Horrible bad luck! During the weekend, we lost a disk on our TSM server. On monday, we replaced this disk and, while it was rebuilding, lightening struck twice and we lost a second disk, thereby causing the wholesale murder of our RAID5 file system.



As a consequence, I have the honour of rebuilding this server from scratch. As a further consequence, this is the first time that I have ever performed such an operation.



I have, as you might expect, a few million questions. To begin with, I immediately placed a software service call to IBM (in the vague hopes that they can somehow assist with the procedure). They have assigned me an individual who is on standby until the new hardware arrives. So, I've got that going for me.



Step 1, after the hardware arrives, will be installing a basic Linux operating system (same RedHat release as the previous incarnation - same kernel version which is known to work with the latest IBMtape drivers). During this procedure, I believe I need to make absolutely certain that I create the exact same partition structure as the original machine (that is to say, same number of partitions, although the exact size of the partitions doesn't really matter, so long as they are big enough to restore all of the files from the most recent backup). Is this correct?



Step 2, after I have an OS, I need to recover the DRM recovery plan file which is stored at an offsite location. This should contain priceless information such as volume history, device history, server options, that sort of thing. Is this correct?



Step 3, I will install TSM server (base 5.2 and then install patches so that it matches the exact same version as it was before the destruction). Most likely I need to install the licence information at this time too. Then I will prepare the database and recovery log file systems. Then I will make certain that I can communicate with the tape library. Then I will mount the most recent database tape and perform some kind of database recovery operation. Is this mostly correct?



Step 4, I will install the TSM client program and restore my server operating system back to it's most recent incarnation. Is this correct?



I guess I am mostly begging anyone out there: if you have performed such an TSM operation during your career and you have any advice or the basic template for accomplishing this small miracle, I wouldn't know how to express my gratitude.

 

[PJ]

Bad luck.

1.) Yes. You've got to install the basic Linux and devicedrivers (ATape)

2.) All you really need is volhist and devconfig

3.) License is not really required at this point. Provided you've got the volhist file, what you need is enough space for the db and log volumes. Format a server of the appropriate size, make sure the library is reachable and run dsmserv database restore.

4.) I don't know what else you run on your TSM Server besides TSM, but I don't think you can "overwrite" the running OS while the TSM server is up (which it must while you're restoring). But at this point you should be up and and happy again anyway.



Cheers

PJ

 

[Harry_Redl]

Hello,



not much to add to PJ's post ... it will work



However I have another procedure I use especially when I do not have enough information about environment I am trying to restore ...



1) install OS and device drivers

2) install fresh TSM and start it

3) in this new instance define (using standard TSM commands) DB and LOG volumes to have enough space for DB restore, extend DB and LOG

- volumes do not need to be on the same place as original ones

4) define the library, drives and paths in this new instance + corresponding device class

- it is VERY likely that with new hardware your device mapping will change

5) backup devconfig and volhist and then halt the TSM



What will you get using this?

working dsmserv.dsk and corresponding DB + LOG volumes

working devconfig out with current library, drives and paths definitions

almost useless volhist.out

default dsmserv.opt - good in this moment



Now all you need is to know on which tape (file) is the DB backup and know its type (DBBackup, DBSnapshot)

Put the tape into library, modify the devconfig out (just the inventory section) and run



dsmserv restore db devc=<device_class> source=<dbs/dbb> volume=<volume_name> commit=yes



notice that this way you do not need the original four "magic files" nor the DRPlan - just to define "bigger" TSM DB and LOG and know the tape with the DB backup ... (and even that can be found by scanning the tapes)







If this is too confusing or you have your DRM Plan file AND you know how to use it - stick with DRM plan ....







Hope it helps



Harry