Serviceguard for Linux

[James Tann]

Hi,

I am trying to set up a 2 node serviceguard cluster on RedHat Linux 3ES (AMD 64 bit) and get an error regarding saying that cluster lock lun is different between the nodes.

Error: Physical lock lun device "/dev/sdb1" on node "luxmqpri" does not correspond to the same device as the physical lock lun device "/dev/sdb1" specified for node "luxmqsec". It cannot be used for cluster lock.
Error: Invalid data for cluster lock LUN configuration

I know this is the same device, there are only 6 EMC disks and it is the only available disk that has this disk partion.

I have setup the small partition in Linux (mode 83) and confirmed this from both nodes.

Has anybody seen this before?
Thanks very much.
James.

[Steven E. Protter]

Shalom James,

The lock lun(disk) is a disk that decides contention between SG nodes when heartbeat is lost.

It must be the same on both systems.

There are several common causes. The most common cause is the cluster configuration files in /etc/cmcluster are not the same on both nodes. Check this and then try cmcheckconf/cmapplyconf once again.

The lock LUN(disk) must be on shared storage.

SEP

[James Tann]

Hi Steven,

The cluster config file is the same both sides and the lock disk is on a shared EMC array.

What are the other causes that you know of?

Thanks for taking the time to reply.
James.

[IT_2007]

James,

Make sure that you see same device name from both servers. For some reason, device name is not same.

[Kodjo Agbenu]

Hi James,

Have you got this ?

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1033536

May be helpful.

Good lcuk.
Kodjo

[James Tann]

Kodjo,
I have seen that but thanks for sending it.
The device file is definitely the same.
James.

[David Child_1]

James,

In case you haven't done so, try running;

/usr/local/cmcluster/bin/cmscancl -o /tmp/cmscancl.out

Note: it uses 'rsh'. On my servers only ssh is allowed so I modified the script and replaced all occurances of rsh with ssh.

This will give you all kinds of details about the environment and what both servers see. Perhaps it will show you where the problem is or maybe you can post the output.

[James Tann]

Thanks for your mail David but I found the cause of the problem last week and had forgotten to close the thread (oops). I was pointing to the correct disk but the EMC array had a setting to send the channel number back to the server. The probe used by serviceguard saw the disk as 2 seperate disks although LVM considered them to be the same disk. A change on the EMC array and all is good in my little world.
James.