Clustering issues

[Kenneth Holter]

Hello all.


I've set up a two node cluster for Apache high-availability testing
purposes. The nodes are connected by the eth0 device.

When issuing a "ifdown eth0" on node A simulate a link breakage, node B will
start fencing node A and fill up /var/log/messages with "fencing node
A...fence failed" messages. The service is not relocated to node B even if
node A remains down/offline. As far as I can understand, node B is waiting
for the fenced daemon to return successfully before it starts the httpd
daemon.

If this is so, what can I do to prevent this from happening?

And another issue: We're not going to be using any shared storage in our
cluster, so which fencing method should I go for? We're not using network
power switches either.


Regards,
Kenneth Holter