How to prevent Serviceguard mounting disks that have been manually mounted

[Mark Bly]

Hi,
We recently had a hardware issue with a server in a 3 node cluster which ended up with us having to rebuild the server from Ignite.
During this recovery we had to stop the Serviceguard cluster as we experienced a TOC on the remaining 2 nodes.
The shared disks and IP addresses were then manually configured / mounted on one of the 2 remaining nodes. (after doing a vgchange -c n)
We are now in a position to restart the Serviceguard cluster and add the packages back one at a time.
My questions are:
1. How do we stop Serviceguard from trying to mount the packages automatically when it starts? (cmruncl)
2. As we have done a vgchange -c n on the volume groups, will this prevent Serviceguard mounting the volume group (we could use this to prevent Serviceguard running the packages that are still manually mounted)

This is a 24 x 7 production system and we only have downtime for 1 package at a time. Therefore I would appreciate any other comments on how we can return to Serviceguard in a controlled mannner.

Thanks for your help.
Mark

[Prasanth V Aravind · Accepted solution]

Hi,

What is the status of cluster now ? running or down???

If it is stopped & want to run this ...

Start the cluster with cmrunnode command not cmruncl... Normally cmruncl starts packages too..

If you run cmrunnode on all nodes they will start cluster daemons & from cluster..

After this you can start individual packages on desired node.

Ensure that you have made the pkg vg in cluster aware before starting pkg manually.

Gudluck
Prasanth

[Prasanth V Aravind]

As you done vgchange -c n .. the vgs won't be getting active onnext cluster pkg start..

You have to do

#vgchange -a n
#vgchange -c y

to make vgs cluster aware.


Gudluck
Prasanth

[Aneesh Mohan]

Mark,

I belive you have to set AUTO_RUN=NO is the
package configuration file .

In that case you should run cmrunpkg manually to activate the package/resources.


Regards,
Aneesh

[Mark Bly]

Hi,
Thank you for the quick reply.
The cluster is currently down.
We usually do the cmruncl (sometimes cmruncl -n if we only want some of the nodes up, then cmrunnod to join the other nodes to the cluster)

I will try the cmrunnode on each server as you suggested.

Mark

[Prasanth V Aravind]

Ensure that you run cmrunnode on atleast 2 nodes.

execute cmrunnod on single node won't start cluster. it will wait for another node to join

Gudluck
Prasanth

[stephen peng]

just agree to Mohan, disable package AUTO_RUN, and then cmruncl, stop package application and unmount filesystem, de-activate vg, vgchange, cmrunpkg, after all package startup, enable package AUTO_RUN.

[Benoy Daniel]

touch a file called debug under /etc/cmcluster/.