"backup" command causes high swap/paging

[kozlow]

When I run backup on a file that is 180Gig, my system goes into heavy paging/swapping. This causes the DataBase (DB2) and application to become unusable.

The file is a database backup to disk. I am trying to get it to tape so it can be sent offsite for DR.

We where breaking the backup into tablespace backups to disk and then coping the files to tape. Each was at most 64Gig. DB2 will not let you recover a tablespace backup to another server.... I can't figure why, but that is what we found out when trying to test a recovery...

Any ideas on how to get the backup to tape without causeing the application and database to choke?

 

[mrn]

The paging is caused by caching it's a "normal" process during copy/backup. There are some mount flags that can prevent this but that would mean stopping your app & Db unmounting the disk then remounting. If you have enough disk space you could try using snapback (see man snapback) or pinning the memory used by DB2; but this would slow your backup.

Mike

Unix *is* user friendly. It's just selective about who its friends are.

 

[mbourne2006]

What about tuning VMM to favour working set rather than file cache?

What sort of values do you have for maxperm% & minperm% ?

 

[DSMARWAY]

how much physical memory do you have in the server ?
how many users connect to the server ?
how big is your swap space?

 

[jstacey]

I second the VMM tuning.

You can try setting

minfree to 120 x #of CPUs
maxfree to 128 x #of CPUs
minperm to 10%
maxperm to 20%
maxclient to 20%

and see how that works in your enviroment.