We have a NW 6.0 SP5 2-node cluster that has been running relatively
problem-free for nearly five years. The hardware is two DL380 G-2 servers
which are identical except that one has an HP gigabit NIC (Broadcom,
really, I think -- it uses the Q57 driver) taken from a DL380 G3 and the
other has an Intel gigabit server NIC. These NICs were put in over a year
ago, maybe over two years ago.

Last August, the one with the Intel card began to abend on random weekends,
typically when the backup ran against the cluster volume. At first it was
irregular, but by November it had gotten to be every weekend. I shifted
backup times and the abends stopped temporarily. It still abends on
weekends, but not when the backup is running.

This morning we discovered that the cluster-enabled volume hosted by this
node was worng. Console1 didn't show any problems, but Windows Explorer,
and any other application that used it could not see this volume. Browsing
to the virtual server showed it was empty, although the volume was mounted.
The non-clustered volume object was OK, though. Migrating the volume didn't

Later I discovered that Console1 reported errors when trying to view the
cluster volume's properties. They went like this:

ERROR: -321 from GetServerConnection


Unable to get the attributes of the object.

Unable to get the details for the selected object.
Some pages may not be available.


Failed to find a serve which handles the volume Cluster1_Files.Mason. The
returned error code
was -321.


Failed to get the current label on the volume. The returned error code was

Then I got the properties page, although it was missing some tabs.

The General/Server Information tab on the virtual server object's
properties claims that time is not in sync, though dsrepair says that it
is. When I saw this, I ran dsrepair against the local database (no replica
on this server). (Not Unattended Full Repair, but Advanced, etc.) The
server abended.

After it came back up, I unloaded NCS and ran dsrepair. It found one error
in Rebuild Operational Schema. I then loaded NCS and ran the repair again,
with no errors and no problems.

It looks to me as though the cluster volume object is corrupt. I know I can
delete and re-create this object, but has anyone got any insight into this?

Incidentally, I looked in Novell's patches for any recent cluster-specific
patches and found none, but I have no confidence in my ability to find
files on the "new" Novell web site. Are there any?

Richard White
Mason County, Washington