topic Re: MC/SG Cluster in Operating System - HP-UX

MC/SG Cluster

kholikt — Wed, 26 Oct 2005 04:13:39 GMT

I have one cluster running oracle database. However, I need to do some maintenance to change some oracle parameter. If I stop the cluster package I won't be able to see those clustered aware file system. I do not want the cluster to failover. I remember that I can some how create a debug file so that the cluster will not failover. Any idea?

Re: MC/SG Cluster

Bharat Katkar — Wed, 26 Oct 2005 04:29:58 GMT

If you want the VG that are active in cluster to be available after cluster is put down then:
1. Before you halt the cluster execute:
# vgchange -c n
2. Halt the cluster
3. # vgchange -a y
Mount the filesystems.

Again while adding this VG's back then:

1. vgchange -a y
2. vghange -c y
3. vgchange -a n

Then start the Cluster.

It should work.

Hope that is what you are looking for.

Regards,

Re: MC/SG Cluster

Prasanta Kumar Bindhani — Wed, 26 Oct 2005 04:31:46 GMT

Hi
you can halt the package
and start mount the filesystems in exclusive mode.

Example :
#cmhaltpkg -v
#vgchange -a e
then mount the filesystem.

After that you can proceed for the S/W level changes.

Hope this may help you.
With Regards
Prasanta

Re: MC/SG Cluster

RAC_1 — Wed, 26 Oct 2005 04:32:03 GMT

you can disable package switching as follows.
cmmodpkg -dv "package_name"

Then you can stop the package. Do vgchange -a e vgxx on cluster aware vgs and mount. then do changes, unmount and deactivate the vgs. vgchange -a n vgxx

then start pkg-cmrunpkg, enable package switching-cmmodpkg -ev "package_name"

Re: MC/SG Cluster

kholikt — Wed, 26 Oct 2005 04:39:38 GMT

I remember we can touch .debug file to some directory then the cluster will be in debug mode.

Re: MC/SG Cluster

Bharat Katkar — Wed, 26 Oct 2005 04:44:51 GMT

sorry ..ignore my post ... i thought you are bringing down the cluster...

RAC pointed you correctly.

Regards,

Re: MC/SG Cluster

Hanwant Verma_1 — Wed, 26 Oct 2005 04:50:40 GMT

Check the following docs

Testing Cluster Halt and Reconfiguration

http://docs.hp.com/en/B5158-90001/ch06s02.html

Hanwant

Re: MC/SG Cluster

Gary L. Paveza, Jr. — Wed, 26 Oct 2005 06:49:52 GMT

I believe you're talking about the situation that happens when you have an oracle database and a service in the package that monitors the database. If the database goes down, then the package fails. Here is the modification to the oracle script. Once this script is running as the service (monitor parameter), all you need to do is touch /var/tmp/${ORACLE_SID}_MAINT to temporarily turn off monitoring. Remove the file to resume.

"Edited to comply with ITRC Guidelines, post of purchaseable scripts not allowed"

Re: MC/SG Cluster

Frank de Vries — Thu, 27 Oct 2005 02:12:03 GMT

I did not see this point confirmed which in fact was your question.

When you touch the debug file in /etc/cmcluster
it will restrain the Mc/Cluster from starting up after a boot.

You can at any moment of course halt a
particular package or de-tach a particular node from a cluster without it to failover.
(This could be handy if you want some part
of the cluster to stay active or accessible)

You can also stop the whole cluster,
and do the vgchange bits to get our
mounted volumes back online.

But you have many technical examples on this from other forum-members.

I just thought it proper to put the debug
technique in perspective.

Hope this helps,
regards,

Re: MC/SG Cluster

unixdaddy — Thu, 27 Oct 2005 05:28:50 GMT

The debug flag is available if you use the serviceguard-developer-toolbox.tar which contains the serviceguard-toolbox-template.tar which allows you to setup serviceguard package using a general template (like the Master toolkits). The configuration file states:-

"###############################################################################
# TKIT_MAINTENANCE_FLAG:
# Maintenance flag indicates if the maintenance mode feature of toolkit is
# enabled or disbaled. This flag may be set to either "yes" or "no". When set
# to "yes", monitor script checks for the existence of the file, "haapp.debug"
# in the package directory. If the file, "haapp.debug" exists and
# TKIT_MAINTENANCE_FLAG is set to "yes", then application monitoring is
# suspended, and the application maintenance can be performed without first
# halting the package. To resume application monitoring after maintenance,
# manually remove the file, haapp.debug from the package directory.
# The default setting for this variable in the toolkit template is "yes.
#
# Note: if Maintenance flag is set to "no" then the above feature would not be
# available and toolkit cannot be brought into maintenance mode.
#
# For e.g., TKIT_MAINTENANCE_FLAG=yes
###############################################################################

TKIT_MAINTENANCE_FLAG=yes"

or as an alternative you can use the following link will allow you to do the same:-

How to keep package file systems mounted after a package halt DocId: UXSGKBQA00000030 Updated: 1/24/02 6:31:00 AM

Re: MC/SG Cluster

unixdaddy — Thu, 27 Oct 2005 06:47:52 GMT

BTW your points to questions ratio is very low - 284 out of 780.