Hello everyone:

I have a customer who has been running a 2-node NCS cluster for about
two years without incident.

Recently, they bought a couple of new servers, and decided to
virtualize the existing Netware servers rather than migrating them to
the new hardware.

I installed a Linux host OS on one of the new servers, added VMWare
server (free edition, build 29996), created the virtual machines, and
used Portlock to virtualize Netware server "B". All was OK for about a
day, then server "A" abended with "Ate Poison Pill in SbdWriteNodeTick
given by some other node". Had to manually fail the cluster over to
server "B" to get users up and running. Even after removing the server
"B" node object from eDirectory, removing and reinstalling NCS, server
"B" passes a poison pill to server "A".

I've tried dedicating eth1 to Netware (rather than having both Linux
and Netware using eth0), no luck - can't run both servers. If both
servers try to join the cluster, server "A" takes a poison pill. I've
checked IP communications in both directions, OK. Did a port scan to
ensure the host OS didn't introduce iptables or other firewall, ports
(including the normal NW ports 427 and 524) are open and listening OK.

Setup details are: Ubuntu server 6.10, 64-bit edition host OS, Netware
6.5 SP5, NCS 1.8. Using the LSI SCSI driver, and CNEAMD as the NIC.
The servers are sharing an external SCSI RAID array.

I've already called Novell and found that they don't support NCS in a
virtual environment. Can someone here offer some insight?

Much obliged.