redhat linux remove quorum device

[B kiran]

hi, I am using redhat linux 5.3 with 2 node cluster.Trying to remove existing quorum device and add new quorum coming from different SAN.
From existing threads posted here I understood that neither cluster nor server will be effected while performing this task.

after stopping quorumd created new quorum device using mkqdisk command. removed the old quorum from cluster using cman_tool leave remove and ensured the quorum isnt member using cman_tool nodes command. But the clustat is still displaying this quorum as a member.

after editing cluster.conf file with new quorum device tried to restart cman, but got below error. could someone assist me in resolving this issue.

Stopping cman... failed
/usr/sbin/cman_tool: Error leaving cluster: Device or resource busy
[FAILED]

[John Bigg]

Sounds like RH cluster not Serviceguard to me. Sorry, cannot help you here.

[Matti_Kurkela]

> removed the old quorum from cluster using cman_tool leave remove

The command "cman_tool leave remove" does *not* remove the quorum device: when you run this command on a node, that *node* will attempt to leave the cluster. If you have cluster subsystems (like DLM or GFS) running at this point, the command will fail.

I've seen this exact same misunderstanding once before on this forum: what made you think "cman_tool leave remove" was appropriate in this procedure? Perhaps there is an error in some commonly-used document.

You don't have to stop cman to change the quorum device; if your cluster has no other problems, you can shutdown and restart the qdiskd service only.

If you change the quorum device while the cluster is running, the old quorum device may remain listed in the clustat output (but please note that it's state will be "Offline"). Once the quorum device change is complete, you may have two quorum devices in the clustat output; this is a cosmetic problem only and should not affect the functionality of the cluster (or so RedHat's "RHEL 5.4 Storage and Clustering" course instructor told me).

I think more information about the current state of your cluster and node is needed before I can suggest a fix for your problem.

Please show the output of all these commands:

service cman status
service clvmd status
service rgmanager status
service qdiskd status
service gfs status
service gfs2 status
cman_tool nodes

If "service cman status" indicates cman is running, please show the output of "cman_tool services" too.

Perhaps your earlier "cman_tool leave remove" caused cman to stop? In that case, run "service cman start". If that works correctly, verify that all the cluster services you need are running.

MK