[Linux-cluster] Info on quorum rename on rhcs 5.8
Fabio M. Di Nitto
fdinitto at redhat.com
Fri Sep 28 10:35:26 UTC 2012
On 9/28/2012 11:40 AM, Gianluca Cecchi wrote:
> I saw some discussions in the past regarding the subject, such as:
> My config is with an rhcs based on rhel 5.8 two nodes + quorum disk cluster and
> I'm evaluating SAN migration from a storage array to another,
> including data disks and quorum disks, limiting downtime and
> considering pros&cons of several scenarios.
> One of the problems is quorum disk moving.
> I start in a situation where the nodes are seeing both the SANs and
> cluster data disks (in HA-LVM config) are already on new storage while
> quorum not.
> My test approach is this:
> - stop qdiskd on first node and wait for stabilization
> second node becomes master
> - stop qdiskd on second node and wait for stabilization
> - cluster maintains quorum (2 votes instead of expected 3 for each server)
> - format the new quorum lun with a different label
> - change cluster.conf accordingly and update/distribute the config with
> <quorumd device="/dev/mapper/mpnquorum" interval="5" label="nquorum"
> log_facility="local4" log_level="7" tko="16" votes="1">
> - start qdiskd on first node and wait for stabilization
> it becomes master
> - start qdiskd on second node and wait for stabilization
>>From qdiskd point view all is gong well but openais shows:
> Sep 28 11:21:00 noracs1 openais: [CMAN ] unable to re-register
> quorum device: device names do not match
It´s a lot simpler if you do:
- drop <quorumd device="..." and just use label=
(this is a general thing)
- stop qdiskd on the nodes as you also suggested
- format the new qdiskd device/LUN with the current label
- format the old qdiskd device/LUN with another label
- restart qdiskd
qdiskd will find the device by label and start using the new one
No need for cluster.conf updates, qdiskd downtime is minimal.
More information about the Linux-cluster