HPUX - Serviceguard 11.18 with hpux 11iv2 nodes

[Tom Haddad]

I have an 8 node cluster running sg 11.18
Each node runs it own package with failover as active/standby.

Basic question in regards to configuation:

If I halt a package AND change the AUTOSTART_CMCLD setting to 0 in the /etc/rc.config.d/cmcluster file, When I reboot the node/server, in addition to the package not starting (since I halted it) will the node join the cluster or will I have to run cmrunnode before starting the packages.

Generally if Im working on the server (ie patching), I of course halt my packaage. Is it a safe practice to change the parameter to 0 also?
Thanks

[Matti_Kurkela]

If AUTOSTART_CMCLD is set to 0, the node will not join the cluster automatically at reboot. You must run cmrunnode before packages can start on that node.

Changing AUTOSTART_CMCLD to 0 for maintenance work is safe. Just make sure you remember to change it back to 1 when you're done.

MK

[Tom Haddad]

thanks..thats what I was thinking.

Also.
My first server was designated as my coordinating node (where cluster ascii file was first created).
As a practice I copy the cluster ascii file to all nodes if I make a change. If this coordinating node is offine, my cluster remains up as my quorum server is set for 50/50. If I chose to keep this server down for good, I can alway chose another server as my coordinating one right?
When I first starting using SG i thought the coordinating node was designated sowewhere in configuration..but its not.
Its a matter of my procedure/documentation on how I want to designate my SG 'roles' for each server.

[melvyn burnard]

>If I chose to keep this server down for good, I can alway chose another server as my coordinating one right?

any node in the cluster can be the cluster co-ordinator, you basically have no control over which one gets chosen. If you are lookin at having the one node down permanently, you should remove it from the cluster

[Tom Haddad]

ok..
So as long as my ascii file is keep current on all nodes. I can chose another server as a coordinating node t some point to perform edit/cmcheckconf/cmapplyconf tasks on the ascii file.

[amipankaj]

Yes, and in case you will have inconsistant cluster config files on different node, you will get auto generated alert for and that and you have to fix that (have to make config files consistant) across all the nodes of the cluster.

[Tom Haddad]

Thanks for all the good info.

[Tom Haddad]

Good info... went forth and reviewed my cluster ascii file and confirmed that all nodes has current file..