Hello there

Recently we added a third node to our Netware cluster to provide better
protection against Split Brain conditions and load balance when a node
is down due to failure or maintenance.

Today we did some tests to check some failure scenarios.

So we started to pull network cables, Fibres from SAN, rebooting FC
switches etc.

All test went fine (with predictable results) except one: When we pull
out the network cables from the node which owned the Master IP Resource
the other 2 nodes abended with SBDWriteNodeTick and the isolated node
loaded the resources (!). And second try gave the same results. Just the
inverse of the expected behavior.


Netware 6.5 SP5 with multiple patches except this one
http://download.novell.com/Download?...d=yBHwgdjigA8~ because is
marked as Field Test Beta.

SAN: IBM DS4300 and Brocade FC Switches. 4 paths to storage server.

Network: 2 nodes are attached to our network core (Dual 3Com Switches
"clustered" with XRN) a third node is on a IBM Bladecenter attached to
the core network through it's Nortel switches. All nodes are connected
with dual nics teamed with Broadcom's BASP software (SLB mode).

The aforementioned patch notes talk about a similar scenario with two
node cluster and suggest as workaround adding a third node. That's why
we decided not to patch the server

Do we have to apply the patch? I find this very strange, should be very
easy for NCS to arbitrate the isolated master with three nodes and it's
not working at all.

Note: When we isolated a non-master node all worked fine.

Thanks in advance and sorry if there are errors in my spelling or grammar.

- Vctor