subnet mask change in cluster

[Vball23]

On my 2-node cluster, i have 2 lan cards direct-cabled between the 2 servers for heartbeat purposed. They are 10.1.1.27 and 10.1.1.28. Recently a remote part of our company built a new computer lab and made all their IP's 10.x.x.x. So, when they tried to connect from there to these 2 servers, the heartbeat cards advertised that they were supposed to take their traffic, instead of the front-office cards. To rectify this, i had to change the subnet mask of the heartbeat cards to 255.255.255.0 and I made this permanent in the netconf file.

Now, I can't make changes to my cluster because of the change in netmask and get the following error:

egin cluster verification...
Warning: Volume group /dev/vgblcopsp01 is configured differently on node ux27 than on node ux28
Warning: Volume group /dev/vgblcopsp01 is configured differently on node ux28 than on node ux27
Error: The subnet for interface lan2 on existing node ux27 has been changed from 10.0.0.0 to 10.1.1.0.
Error: Changing the subnet on an interface that is configured in ServiceGuard is not supported when the cluster is up.
Error: Use ifconfig to reconfigure the old subnet or halt the cluster and re-apply the configuration to use the new subnet.
cmcheckconf : Configuration check failed.

I've stopped the cluster and restarted it, but I think I need to put that netmask change in a config file. If so, where?

[G. Vrijhoeven]

Hi,

You should put that in /etc/cmcluster/cmclconfig.ascii ( Cluster ASCII ) file

What you can do is cmgetconf -v -c /etc/cmcluster/cmclconfig.ascii

Adjust the network settings, cmcheckconf and cmapplyconf -v -C /etc/cmcluster/cmclconfig.ascii

HTH,

Gideon

[John Poff · Accepted solution]

Hi,

As Gideon mentioned, you'll have to to a check/apply on your cluster ASCII file to pickup the changes to your subnet mask. The subnet mask isn't configured in the cluster configuration, but the cluster detects it.

Here is a link to latest Managing ServiceGuard manual:

http://docs.hp.com/hpux/onlinedocs/B3936-90079/B3936-90079.html

Take a look at the chapter on Cluster and Package Maintenance, in the Reconfiguring a Cluster section. It has a nice chart that shows the various cluster changes and the state the cluster must be in (up or down) for each change.

JP

[Sanjay_6]

Hi,

You need to do a cmquesrycl once again, edit parameters in the ascii conf file for the cluster, do cmcheckconf and then cmapplyconf to apply the changes. On the surface the ascii cluster conf file may not have changes, but you still need to reapply.

Hope this helps.

Regds

[Geoff Wild]

Yes, you do have to re-run cmquerycl, cmcheckconf, and cmapplyconf.

In "Reconfiguring a Cluster " section of "Managing Serviceguard Version A.11.16" it states "Change IP addresses for heartbeats or monitored subnets
Cluster must not be running."

http://docs.hp.com/hpux/onlinedocs/B3936-90079/B3936-90079.html

Rgds...Geoff

[G. Vrijhoeven]

Hi ( again),

you might want to adjust your lvmconfiguration as well:
Warning: Volume group /dev/vgblcopsp01 is configured differently on node ux27 than on node ux28
Warning: Volume group /dev/vgblcopsp01 is configured differently on node ux28 than on node ux27

An vgexport on the primary node && vgimport on the secondairy node will illiminate the Warning.

Regards,

Gideon

[Vball23]

Thanks, that's exactly what I was looking for.

Also, I have been cleaning up the lvm stuff, but thanks for the heads up. Apparently one of the fiber paths was moved while I was out and when it was re-added, hardware paths changed.

[Vball23]

Solution found. RTFM!