Oracle startup script in m/c service guard

[naveen_5]

I 'm using oracle 8.1.7 on m/c service guard a.11.13 on n-class servers.
Oracle is installed on volume groups in Fc-60 disk array.
Therefore for package move( package switching on failover), Vg's containing all of oracle including data, are removed
What is the correct way to implement shutdown/startup of oracle in such a way that the Vg's are there and to be able to issue the commands.
Does B5139BA have different package control script, to aviod this?

[Printaporn_1]

why worry , if oracle binary and start script is under /u01 in primary node.
when the vg switch to adoptive node , it's mount under same name /u01 and the same package control script still can be use.

[Steve Edmunds]

That's not what Naveen meant.
Can someone explain the default MC/serviceguard logic better
so we can understand how to script, and what to script for
a package containing Oracle with files on a volume group that will
move - to start/halt properly. By default is the behaviour of the
scripts to run vgchange AFTER shutting down the databases during pkghalt on first node, then conversely, to run the vgchange BEFORE the start
script on the new node - so that in both cases the files are available?

At the moment it seems vgchange just gets run first in either case,
is there a variable to set to enable the above logic? Or are we
missing something?

Many thanks!

Steve

[melvyn burnard]

The Enterprise Cluster Master Toolkit (B5139BA) does have sugeested scripts for Oracle, and other databases as well.
You can if you wish use standard package scripts to do the work, but these are more intuitive to set up, and also include a monitor script to monitor the Oracle processes themselves.
As for explaining the logic, they are self-explanatory if you read through them, as indeed any standard control script is.
These scripts call various sections one after the other, hence you will find the vg activated before you mount a fs, and the fs mounted before you call any commands that are on the fs.
And the reverse is true, just go read the start/stop areas of the control scripts.

[Steve Edmunds_1]

Thanks Melvyn,
We did indeed figure it out
by looking at the scripts, and
we do want to monitor properly so we will get the cluster toolkit.

[Justo Exposito]

Hi,

I am using the standard scripts for startup and shutdown Oracle in my cluster without any problems.

Regards,

Justo.