Updating OpC.cntl file. Does MC/SG need to be updated too?

[Gino Castoldi_2]

Hi,

Mgmt Server: ITO 5.39, Oracle 8.0.5, NNM 6.01, MC.SG 11.09, HP-UX 11.0.

We have a two node cluster that the ITO Mgmt server runs on.

We need to update the /etc/cmcluster/OpC/OpC.cntl file because we are upgrading Oracle to a new release 8.0.6

In the "OpC.cntl" file it has the path for
Oracle listed in several areas: ORACLE_HOME=/u01/home/dba/oracle/product/8.0.5
which we need to change to (8.0.6).

Do we have to make any other changes other then updating the OpC.cntl file?
Do we have to update the binary cluster configuration files themselves?

(10 points to any good answer).
Thank you.
Gino.

[Dirk Wiedemann · Accepted solution]

Hello Gino,
don't worry, MC/SG dosn't need to be updated. If you only change the .cntl-file, you only have to distribute the new .cntl file to every node in the cluster.
And - of course - you must check the starting and stoping of the related package.
regards
Dirk

[Gino Castoldi_2]

Hi,

Thank you for the reply. Does that mean we need to update cluster configuration files?
(i.e. run cmapplyconf, etc)?

(10 points to any good answer).
Thank you Gino.

[Sridhar Bhaskarla]

Hi Gino,

You do not need to re-apply the configuration. Changes in control file do not need to reconfigure service guard.

-Sri

[Gino Castoldi_2]

Hi,

I am a newbie when it comes to MC/Service Guard. Can anyone tell me the basic procedures we would need to take in order to make the change to the cluster ascii file?

Thank you again.

(10 points to any good answer).
(I'll assign points in a short while, I'm just looking for more information at this point).
Thank you Gino.

[Jeff Schussele]

Hi Gino,

Yes - IF the ORACLE_HOME changes, you'll need to change the control files on all nodes.

I'ts a good idea to define the ORACLE_HOME to a link to the actual home dir & then all you'd need to do is change the link - something like

ORACLE_HOME=/u01/home/dba/oracle/product/

Then /u01/home/dba/oracle/product -> /u01/home/dba/oracle/product/8.0.5

can just be changed to

Then /u01/home/dba/oracle/product -> u01/home/dba/oracle/product/8.0.6

Rgds,
Jeff

[Sridhar Bhaskarla]

Hi Gino,

I recommend you to check the following document on Service Guard.

http://docs.hp.com/hpux/onlinedocs/B3936-90053/B3936-90053.html

The section "Cluster and Package Maintenance" will be of interest to you.

As per your previous question (on behalf of Dirk), the common error we do while doing the maintenance is forgetting to distribute the changes to the other nodes. Make sure that your ascii and control files are the same on all the nodes.

Also, before making changes to control file, make sure the corresponding package is down. It is not mandatory, but there is a good chance that the running package may not understand the changes and may get into trouble while coming down using the new control script.

In general, modifications to the control file do not need re-applying service guard configuration. Some changes for the cluster and package ascii files can be done with the cluster|package up and running. You can see more details in the above document.

-Sri

[Dirk Wiedemann]

Hi again,
ok, some basics:
to create a new ascii file with the actual settings you can use the command
# cmquerycl -v -C /etc/cmcluster/cmclconf.ascii -n nodename -n nodename ...
In the ascii-file you can configure various settings for volumegroups, networksettings and so on.
Once you have configure everything you should verify the configuration:
# cmcheckconf -k -v -C /etc/cmcluster/cmclconf.ascii
If everything is ok you can distribute the binary configuration file:
1. activate the cluster volume group
# vgchange -a y /dev/vglock
2. generate binaries
# cmapplyconf -k -v -C /etc/cmcluster/cmclconf.ascii
cmapplyconf creates and also distribute the binaries to all nodes of the cluster
3. deactivate the cluster volume group
# vgchange -a n /dev/vglock
4. finaly you must save all volume group and cluster lock configuration data with the command vgcfgbackup, e.g.
# vgcfgbackup /dev/vg00
use vgcfgbackup for ALL volume groups!

Now the cluster is configured. run various tests (like starting/stoping packages, nodes and the whole cluster).
regards
Dirk