We have a two node cluster running on NW6.5 SP3. One node is poisoned due
to a momentary loss of network connection and abends. The failed node was
restarted. Cluster Status shows that the one and only resource had not
migrated to the other node and was comatosed. The resource was off-lined
and then on-lined back to the newly started node and normal processing
commenced.

This node has abended on a previous occasion (poisoned) yet its resource
failed over without incident. A clue is found on the other node's logger
screen which reports "CLUSTER WARNING 10310 - RESOURCE POOL1_SERVER HAS
BEEN PUT INTO COMATOSE BECAUSE IT MAY CAUSE SERVER ABEND"

The only doc found on the subject was TID 10092849 which attributes the
problem to CRM.NLM, scheduled to be fixed in SP2. The TID explains that if
a node in a two node cluster fails, and its ip number is higher than that
of the other node, then its resources will not fail over, but will
comatose instead.

The node that failed in our cluster had an ip address that was lower than
the other node, and both nodes were running SP3.

Has anyone seen this sort of thing before? We need to find a fix so that
it does not happen again. Would the addition of a third node eradicate the
possibility?

Thank you

Philip Zelazowski - IT Project Team - LGC Ltd.