Hi there,

I posted this in the windows clustering group as well, thought it was
relevant to here as well though.

I have a Windows 2000/Sql 2000 server cluster and while testing it
discovered that if Node 1 is in control of the resources and the private and
public network cables are pulled to initiate a failiure, the cluster cannot
recover.

Is this right?

I would have expected that Node 2 would say I cant see Node 1 on either
network and take the resources to provide the service?

Should it act like this or do I have a problem with the cluster?

Thanks In Advance,

Andy

IF you take the private network cables out
then the heartbeat is not operational
and node 2 will not take over control.

Its a desired behaviour.

Andy,

The way the failovers work is a little more complicated so
that the 'split brain' scenario is prevented where both
servers are trying to take control. The process of failing
over has three steps.
First the heartbeat signal on the the dedicated cluster
network is checked.
If there is no response then the second check is over any
(or all?) of the networks that have been set for 'mixed'
traffic i.e. private cluster traffic and public traffic.
If this still doesn't get a response then the server will
attempt to gain control of the shared disk using a
sequence of SCSI bus resets and reserves. If the other
server removes the reserve and hence keeps control of the
disk then no failover will occur - this is what happens
when you pull out both network cables but the disk is
still connected.

The only ways to check/force the failover is to kill the
cluster service, remove the network and shared storage
links, or just switch the server off!

Hope this helps
Kevin