Unexpected cluster behavior

[José Enrique González]

Hi:

I am running a 2-node Oracle database cluster with MC SG A.11.14 and HP-UX 11.11. Each node in this cluster hosts a package which runs an instance of the main production DB.

In adition to this, on one of the nodes, a RMAN database instance is running. This RMAN instance is NOT defined as part of the cluster. It has no packages associated with it, so it is not monitored. Even its datafiles are ordinary files stored in local filesystems, so they are not in a clustered VG.

However (here is the odd) any time the RMAN instance is shut down to be backed up (4 times a day, through crontab) syslog in the two nodes starts reporting "Updated file /etc/cmcluster/cmclconfig..." as if one of the cluster DB instances had been shut down, until RMAN is brought back online.

Any explanation for this unexpected and so rare cluster behavior? Any help will be truly appreciated.

Jose Enrique Gonzalez
Caracas, Venezuela

[Steven E. Protter]

Shalom Jose,

Explanation,

Someone other than you has configured rman into the clsuter.

The cron job is restarting it regularly.

You are misinterpreting the results.

I'd examin all the package scripts and see which one is starting rman.

Most likely its in /etc/oratab and some process is reading through there and starting it when other database instances fail over or are monitored and started.

SEP

[Srikanth Arunachalam]

Hi Jose,

You may be using the RMAN catalog database and when you are shutting down this instance, are the oracle mandatory processes shuts down ? Your oracle monitor script in serviceguard may be looking at the availablity of the oracle processes, that is getting changed.

I think the SG monitor script should be modified to have a flag in it (unix env variable), that should be controlled by your cron job backup script to stop monitoring at the time of taking backup.
If the flag is set..dont monitor. If the flag is unset,continue with the monitor process.

Thanks,
Srikanth