Re: SAN's, snapshots and online backups questions...

Anton Buijs wrote:
>Bruno Jargot <see_at_reply.to.invalid> schreef in berichtnieuws
>i7rghvchh5ba0okkgrmjjhjejd3anmgoea_at_4ax.com...
>| On Fri, 18 Jul 2003 21:19:54 +0000 (UTC), Daniel Nichols wrote:
>|
>| - Backup (and restore) can be done with the SAN (lan free backup) very
>| quickly.
>|
>| SAN can be expensive but the gain is real. Better availability can be
>| reached.
>|
>| For the backup scripts, I've just developped a script for umounting
>| filesystems from the backup server, mounting filesystems onthe backup
>| server, and launch the copy. The Oracle backup scripts were almost
>| unchanged.
>|
>| My 2 cents,
>
>Very interesting. We are talking about something similar but our Unix system
>managers say it is not very easy to mount disks from the SAN on another
>server. We have a Compaq SAN (I think) and servers running HP-UX 11.
>What software are you using to be able to do all this you describe? What
>platform?

We have a IBM ESS F20 storage server. The servers are running AIX 5.1 We're doing the backup with TSM which allow lan free backup.

For mounting the database on the backup server. We're using Flash Copy (ESS tool for snapshot) :
-> The database is on a production server -> The disks for the backup are attached to the backup server

The drawback is that we need two times the volume of disks (one for the production server, one for backup server)

By default, the backup disks are not mounted on the backup server. From the production server, we launch the FlashCopy (the copy is internal in the ESS) (No need of umounting the filesystems on the production server as Oracle pass through the buffer of the OS).

Then we're mounting the backup disks on the backup server (importvg and fsck) and the backup can start from the backup server.

The "backup" disks are never mounted on the production server. The copy is internal to the ESS. The production server is not aware that a copy is being done in the ESS.

I agree that mounting the disks on the backup server is not very easy but with the great help from an IBM redbook, I was able to write the scripts without problems. ( In brief, I work with the PVID of the disks which are the same on the production server and the backup server. I take the list of the PVIDs from the production server, copy the list on the backup server and use it to know which disk belong to which volumegroup. Once I've identified a disk for each volumegroup, importvg, fsck and mount).

Note : Your posts aren't on my newsserver. I've found it on google. So, if I can help you, email me a copy of your post. Received on Sun Jul 20 2003 - 14:12:34 CDT