Extended Cluster ServiceGuard

[FrankBurg]

Hi everyone. I need to create a serviceguard package to protect hp Dataprotector. I have 2 different sites, each node has hp-ux 11.23 installed, the replication data: mirror-ux. How can I create that package, and what kind of limitations/restrictions can I find?

Thanks.....

[Steven E. Protter]

Shalom,

You will need to create this yourself.

I have a question though. Why?

Lets say your system writing a tape goes down during that process. The second node is not going to be able to take over and continue writing to that tape. The backup is going to fail and need to be re-started.

Even if you have a shared tape device like some storageworks system, a node failure is going to require some kind of intervention.

The limitations you will run into is very likely most failover scenarios, unless they occur when the backup is not running will result in a phone call and manual intervention.

I can see many pitfalls here for questionable benefits. If you maintain a DP database on shared storage, you may be able to prevent data loss by going cluster.

The creation of the script is by standard means, you create the package on the command line and configure it for what volume groups its going to activate, what startup scripts will be run and what monitors you want configured.

SEP

[FrankBurg]

Thanks SEP....

[FrankBurg]

Do you think that using Manage of manager (DP) can I resolve the problem? instead of using a serviceguard package?

[Klaas D. Eenkhoorn]

No, you will not solve that problem by using a different interface.
If the node that has the media agent, the node that has the diskagent or the node that has dataprotector cellmanager fails your backup will be broken, whatever interface you use.
Remember they depend on eachother, the media agent, diskagent/intergrationagent and cellmanager.

Only option is to restart the failed session after a failover accured.
I'm not shure but there seems to bee a function for that, restart failed sessions (??).

Advantage of dataprotector in a cluster is when the package of the cellmanager has switched the scheduled backups will continue as planned from that time on, the tape with the broken backup will be ejected, if requested, and will be set aside or reused in append mode. This keeps your Dataprotector database and Backup/Restore functionality accessable at any time.

Kl@@s.

[Klaas D. Eenkhoorn]

Steven,

Yes, after a failure of the cellmanager the backup currently running will fail and be broken.
But, and this is the advantage and fun part, your dataprotector backup/restore functionality wil be continued on an other node in a matter of minutes/seconds.

Else you have to start a dissaster recovery of your cellmanager (!!!!).
I know what i would prefer, rather a broken backup then a recovery of the cellmanager with the chance of dataloss of backup/tape information.

Kl@@s