cman - losing quorate state
Hi all,
I have RHCS cluster which consists from 2 servers and one quorum disk.
I have following issue.
Due to fact that cman sometimes lose connection to quorum disk I decided to change vote count to prevent losing of quorate state. Each server has 2 votes and quorum disk has only 1 vote. Cluster need at leat 2 votes to start provide the services.
Here is appripriate part from my cluster.conf:
<clusternodes>
<clusternode name="volovec.onsemi.com" nodeid="1" votes="2">
<fence>
<method name="1">
<device modulename="" name="dracvolovec"/>
</method>
</fence>
<multicast addr="239.192.0.1" interface="bond0"/>
</clusternode>
<clusternode name="baranec.onsemi.com" nodeid="2" votes="2">
<fence>
<method name="1">
<device modulename="" name="dracbaranec"/>
</method>
</fence>
<multicast addr="239.192.0.1" interface="bond0"/>
</clusternode>
</clusternodes>
<cman expected_votes="2" two_node="0">
<multicast addr="239.192.0.1"/>
</cman>
I thought that this workaround will fix my issue but yesterday I have bad crash of my cluster.
I stop cman on the second node so all services were running the the first one. Then sudenly quorum cman lost contact to quorum device a shutdown all cluster services.
Apr 14 14:47:33 volovec openais[8894]: [CMAN ] lost contact with quorum device
Apr 14 14:47:33 volovec openais[8894]: [CMAN ] quorum lost, blocking activity
Apr 14 14:47:33 volovec clurgmgrd[27743]: <emerg> #1: Quorum Dissolved
Apr 14 14:47:34 volovec openais[8894]: [CMAN ] quorum regained, resuming activity
Apr 14 14:50:15 volovec clurgmgrd[27743]: <notice> Quorum Regained
I still don't know why!!!
Is there any mistake in my cluster.conf? Did I make something incorrect?
Or my next idea is that this issue could be related to cluster rules. Maybe the cluster (cman on alone node) automatically stop all services in case of losing quorum device despite on fact that it has sufficient votes.
Please send me some reply if you have some idea. It will help me a lot.
Thank you very much.
Regards,
Stefan
|