Oracle agent

[Bill_Weldon]

Has anybody had any experience with the oracle agent? We have about 20 databases and are looking for a better way to back up the data than what we are currently doing. Mainly because there is no solid hand off between the RMan process, Netbackup, and the file management for disk space management. Any and all comments will be helpful.

Thanks

[Brice_Avila_2]

Bill,

Sounds like your backing-up Oracle to disk using RMAN, and then backing-up this disk images using NetBackup? I agree that the transfer is cumbersome.

I've worked with the Oracle agent. Basically install the XBSA library from Veritas, point your Oracle instance at this library (how you do this depends on the version of Oracle your running) and then change your RMAN scripts from:

allocate channel type disk;

to

allocate channel type sbt_tape;

On windows the XBSA library is installed when you install the client; On UNIX the XBSA library install is a separate step. There are some example RMAN scripts available when you install the XBSA library, but it sounds like you have your own. Try starting with the "NetBackup 5.1 for Oracle on UNIX System Administrators Guide" at "http://support.veritas.com/docs/268099". Good luck.

Brice

[Bill_Weldon]

Thanks Brice,

I have some more information form our DBA's. They are running RMAN to both backup the database's and also to manage the files on the disks that they are writing to. The big issue they have is there is no solid hand off from disk to tape nor is there a verification that the data has been archived. They want to keep backing up to disk and modify the process where after their RMAN generated files have been written to tape a certain number of times or after the disk space gets to a certain point then they are deleted oldest first. Someone mentioned vaulting as an option however, I am a bit warry of that and would rather see the backup process be set up where the data is written to tape and disk in concurrent processes.

[Jeffrey_Redingt]

If you are not using the Oracle agent and you are running 5.0 you could either use disk staging which backs up to disk immediately and then later to tape after the disk reaches a certain threshhold set by you or you could simply do inline copy to disk storage unit and tape.

[Stumpr2]

We have discovered that restore time is minimized if you use inline tape copy. The reason is that the threads will be dispersed over more tapes and concurrent restores of the threads are then possible. We went from a 10 hour recovery to 3 1/2 hours :)