[Linux-cluster] Halt nodes in cluster with cable disconnect

[Digimer]

On 01/27/2012 01:51 PM, emmanuel segura wrote:
> It's ok like that, the node doesn't has the sleep always gets fenced,
> because when it tries to use the fence device of the other node to make
> the fence take the sleep, sorry for my bad english :-)

That anyone can learn English as a second language is amazing. I think
your English is fine. :)

> if you use a quorum disk you can aboid this problem with master_wins="1"
> in the quorum tags in your cluster conf && and if you wanna info about
> the parameters for the cluster
> 
> man qdisk ; man fence ; man cluster.conf
> 
> I recommend to use a qdisk if you are using SAN or iscsi, but if you are
> using just DRBD, remember drbd has it's own internal fencing

As mentioned, don't use qdisk on DRBD. In a split-brain condition, both
partitions could get the qdisk votes.

> I had experience with drbd and i think work better with heartbeat+pacemaker

DRBD works just fine on corosync+rhcs. The heartbeat project is
deprecated and should not be used in any new projects. You can use
corosync+pacemaker with DRBD just fine though.

> Remember every redhat cluster version has the diferents problems
> 
> http://docs.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/index.html

Every cluster has it's own set of problems. :)

-- 
Digimer
E-Mail:              digimer at alteeve.com
Papers and Projects: https://alteeve.com