July 17, 2013
In these days I figured this issue:
I had a two nodes cluster with quorum disk configured. During my tasks I had to shutdown both systems and boot one for time.
I was very surprised when I saw the first node stop booting with message “cluster unquorated”.
I asked how that was possible because I configured everything following Redhat instructions.
Looking on the web I found this note from Redhat:
” The problem with starting a cluster with one node and qdisk is that there
needs to be an established cluster before the quorum disk can be added to the
cluster. So if you would like to start the cluster as a single node that is a
special configuration which has to be considered…and if you would like to
drop a node for maintenance then you need to configure the cluster as a 3
node cluster with the third node consisting of the qdisk.”
So this is one of the main reasons ( I think ) why Redhat suggests to install and configure cluster starting with three nodes. The other reason is for the issues related having just two nodes without quorum disk configured ( I faced this issue , I saw a infinite fencing from each node !! )
So: If you see the first node stop booting even if it has quorum disk , DON’T PANIC!
In the worst case where you have the other node powered off and you can’t poweron it , do this:
– boot in single user
– delete quorum disk from cluster.conf
– configure two_nodes directive to 1 ( look to the web for the correct parameter and syntax )
– init 3 and you should have 1 cluster node running and quorate!
July 4, 2013
For recovery purpose , yesterday I made and configured a quorum disk on a two nodes Redhat 5.5 cluster.
I had some issue when I would make it visibile to the cluster.
I tried to use system-config-cluster GUI command , but I saw that it was possible configure a new quorum disk ONLY at “new configuration” time.
So , i tried setting up CONGA and adding the cluster nodes, but this failed too because I had one node offline.
At this time i tought about to make a new configuration on my vmware-test-environment and manually adding the entries but i preferred to recover a node, making it cluster aware available and finally using CONGA!
For everyone: REMEMBER TO start the qdiskd service and put it bootable via chkconfig. Also REMEMEBER to check the heuristic min_score parameter value