SAP Slow Restore due to long browsing times

[wickd]

Hi,

I have a NT 4 backupserver with Legato 6.02 that save several SAP systems via SplitMirror backup by brbackup and backint. This backup system is running for some time(1 year). The index size from the backupserver with owns the backups is only 25 MB. Last week I made a test restore. During the restore all look fine. The brrestore starts, and the client starts browsing at the backupserver. This take a long time from 40 to 60 min before files where restored from tape. The next files from the restore had the same behavior.

Have any one an idea ? Please let me know .

Regards

dieter

 

[esserc]

I might be able to help. First of all most of my experiance is with SAP on Oracle on Solaris with Netowrker on Solaris, but Most of the things I have found should apply.

I've worked with SAP on NT with Networker on NT and I hope I never have to go through that again. It's a nightmare! But, good luck.

First things first, make sure you are using the Networker SAP Module 2.0. Not 1.x or 2.1! They have major problems.

Second don't use the split mirror unless you absolutly have to. Your better off doing online backups on the primary mirror. This is because the relocation on the datafiles during a recovery will force the recovery to a parallism of 1 because it must rename each datafile. Yes folks, brrestore is brain dead. (even version 6.10!)

By the way, if your not using the SAP module for your backups and using the split mirror to do cold backups of the file system, you won't have this problem. The only problem you'll have is that SAP and Oracle will not support you if you have a recovery issue.

Either way, this won't fix your immediate problem (will not directly). Your root problem is that the Networker 6.X file index database does not handle a high number of small files well and SAP loves to write lots a small, useless, files. Job logs, trace files, etc. In fact as we speak, I'm recovering a SAP BW system. Its 1TB and about 500,000 files. My index DB is 36GB and 400,000 index files, so I know how important it is to optimize it.

You can improve things be modifying your backup a little. Let me know more about how you do things, and I think I can give you some more direct help.

 

[Barks]

I had the same problem with the indexes but on HPUX 11.0. I received a new 2.1 backint binary from Legato that has solved this issue, hopefully there is one for NT as well. Hope this helps.

Product: NetWorker
Release: rt_linuxnmo31
Build number: vsadjadi-2001-361
Build date: Thu Dec 27 12:33:58 2001
Build arch.: hpux11n
Build info: unknown host:/net/lonestar/space2/vsadjadi/src/rt_linuxnmo31,DBG=0,OPT=-O

Kent