[Pacemaker] Quorum in pacemaker

[Vijay B]

Hi,

I'm trying to set up a three node cluster using pacemaker+corosync, and I
installed the required packages on each node, checked for their network
connectivity so they can see each other, added the required startup scripts
and edited the cluster.conf file as well so it includes all three nodes.

Now, when on the first node, I attempt to start up cman using service cman
start, it times out thus:

vagrant at precise64-pmk1:~$ sudo service cman start

Starting cluster:

   Checking if cluster has been disabled at boot... [  OK  ]

   Checking Network Manager... [  OK  ]

   Global setup... [  OK  ]

   Loading kernel modules... [  OK  ]

   Mounting configfs... [  OK  ]

   Starting cman... [  OK  ]

   Waiting for quorum... Timed-out waiting for cluster

[FAILED]

vagrant at precise64-pmk1:~$

Why is this? Is it because I have three nodes to begin with in my
/etc/cluster/cluster.conf, and so this node expects that the cluster quorum
is 2, and so it should be able to talk to at least one other node? At this
point, I haven't started the cman or pacemaker services on the other nodes.

If this is the case, what will happen when two nodes of the three die? In
case cluster.conf changes accordingly to reflect the new cluster
membership, what if all three nodes are simply powered off and one
rebooted? The cluster will be down, won't it?

What is the best way to get around this? I don't want to set
CMAN_QUORUM_TIMEOUT=0, since as I understand it, the node would then go
ahead and start itself as a cluster without waiting for the other nodes,
and if this causes my service to start up and it is already started/running
on another node, it could cause issues.

Now, I don't know how to configure quorum disks for pacemaker - is it
possible to do this with pacemaker? How does it work? What are the
recommended ways to address the above problem? I infer that if this disk is
configured, the node that grabs the disk first becomes the president of the
pacemaker cluster. In this context, I have another question - does corosync
have its own cluster membership state distributed across all cluster nodes?
If so, I guess quorum is configured at the corosync level rather than at
the pacemaker level?

Apologies in advance if my queries above are addressed in the documentation
already - I felt it would be quicker and more accurate to ask the community
for reliable info.

Thanks!
Regards,
Vijay
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20140626/7df7e851/attachment-0003.html>