serviceguard config question

[likid0]

Hy,

I need to change the cluster lock disk for a cluster, and i have changed the config file with the new devices, and know i am triying the configurartion with cmcheckconf -C ll.ascii -k
and i get some errors, i wanted to know, if you can repair this errors with the cluster on-line, or do i have to wait and stop the cluster.

wopper is configured as the lock vg

Warning: The volume group /dev/wopper on node macdonald has inconsistent cluster configuration data
This can be cleared using vgchange with "-c".
Error: Volume group /dev/wopper has mismatched cluster ids. (1634729984 != 1146102093)
Warning: Volume group /dev/wopper is configured differently on node burgerking than on node macdonald
Warning: Volume group /dev/wopper is configured differently on node burgerking than on node macdonald
cmcheckconf : Unable to reconcile configuration file cmclconf.ascii



this is service guard 11.14 on a 11i system, i wanted to have a clear idea on what to do to get over this problems before i stop the hole cluster, because once i stop i only have a small window to get it working.



Any help is much apreciatted.

Thnx

[Steven E. Protter]

Shalom,

I believe you don't have the configuration files copied to all nodes.

When working on the cluster lock disk configuration you must copy the configuration file itself to all nodes prior to running cmcheckconf/cmapplyconf

You should be abel to repair the errors without taking the cluster offline. You may need to cmhaltnode one node, make corrections and then rejoin.

I've run into situations however where it is simply easier to bring down the whole cluster , reconfigure the volume group export -p the configuration and import it onto the second node.

I've had the course training on SG and dealt with a few production crisis since coming over to this side of the ocean, so I'm fairly confident you can do it either way.

SEP

[melvyn burnard]

The errors seem to indicate there is an LVM issue with this VG.
The cluster lock VG should be in the cluster ascii file, and the vG must be active on the node where you run cmapplyconf.
I would suggest you check the LVM config by doing
strings /etc/lvmtab
on each node and fixing this first.
Then you will need to halt the cluster, then vgchange -c n on the VG, then vgchange -a y, and finally do the cmapplyconf.

As for the comment:"When working on the cluster lock disk configuration you must copy the configuration file itself to all nodes prior to running cmcheckconf/cmapplyconf"
This is incorrect, you do not need to have the cluster ascii file on all nodes.

[Torsten.]

I would advise to run "vgcfgbackup" for the VG for future use after applying the config.

BTW, I'm pretty sure Stevens intention was to write "... you may copy ..." instead of " ... you must ..." - it is really not necessary, but can't damage anything ;-)

[likid0]

I am doing the cmapplycheck in the node with de vg /dev/wopper active, this is burgerking, but all the lvm errors i get are from macdonald.

The error part looks like it could be solved by doing vgchange -c n and vgchange -c in node macdonald?, can i do this with out stoping the cluster.

And i think you have to stop the hole cluster to change the lock disk no?


thnx again!

[melvyn burnard · Accepted solution]

To change the cluster lock definition, you must halt the cluster, and have the cluster lock vg activated at the time you do the cmapplyconf.