steps/procedures to upgrade oracle on SG env.

[Hanry Zhou]

I have a cluter with 2 active servers, and 1 adoptive server. /opt/oracle01&vg01 and /opt/oracle02&vg02 are as shared file systems on these 2 active servers respectively.

We plan to upgrade Oracle 8i to 9i, is there any procedures you can provide me with regards service gurad concern?

and what issues you have encountered?

thanks,

[Bharat Katkar]

HI,
As long as you are not touching SG acii files you don't have to worry about cluster.
Not much familier with oracle but may be your user_defined command could change in control scirpt that should not affect the cluster as well.
Better to wait for more responses to come from Oracle studs.

Regards,

[Sridhar Bhaskarla]

Hi Hanry,

The procedure I follow is this.

1. Shutdown the packages. Enable the volume groups with exclusive option.

vgchange -a e vgxx
mount /dev/vgxx/oracle01 /opt/oracle01

etc.,

2. Upgrade oracle as you do it on the standalone systems. Make sure you will be able to start/stop the database correctly. Once the work is done, unmount the filesystems and deactivate the volume groups.

3. Check your oracle startup-shutdown scripts in serviceguard and modify them according to the new version.

4. Start/stop the package manually and make sure it works on their configured nodes. Failover the packages to their alternate nodes and make sure it works there too.

5. If you have configured monitoring of the processes through serviceguard, you may want to schedule the testing for that part also.

-Sri

[Hanry Zhou]

I only need to perform the oracle upgrade on one node, and this node could be any one of active/adoptive nodes. Is this a right statement?

thanks,

[Steven E. Protter]

Your statement is correct, but don't you eventually want Oracle to failover like any other package. If so, you are eventually going to need it on more than one node.

SEP

[Hanry Zhou]

SEP,

The /opt/oracle is shared file system, and contains all Oracle software, so that was why I stated ONLY ONE node within the cluster (2 active nodes with 2 packages and 2 adoptive node) need to perform the Oracle upgrade.

Right?