I just got done with a 10 hour migration from ZDM 7.0 HP3 to ZDM 7.0
SP1. In the end, we had to abandon the migration and revert back to ZDM
7.0 HP3.

There are serious flaws with ZDM 7.0 SP1 when it comes to imaging
services running on clustered systems. Please conduct off-line
confidence testing on an isolated network before attempting to migrate a
clustered system to ZDM 7.0 SP1.

We are running a two node cluster on OES NetWare 6.5. We were running
ZDM 7.0 HP3 with ONLY the PreBoot Services and Imaging Services. When we
went to apply SP1, everything seemed to go smooth. As a part of the
installation process, the install will update the local workstation, then
the clustered volume, then each node. It was on the node updates that
something unusual occured.

A java-based login dialogue box appeared and wanted eDirectory
credentials for the IP address of the node. The problem was that there
was no was to enter the information. All of the fields were locked and
none of the buttons (OKAY, CANCEL, DETAILS) were clickable. The only
thing that would work was to click the X to close to dialogue box down.
Then the installation would proceed and the dialogue box would once again
appear for the other node.

When the installation completed, the log files indicated complete
success. No error messages at all. However, after rebooting the
cluster, the following problems cropped up:

- Server Policy Package - Imaging Server rules were being interpreted by
PXE just fine, however, after booting the linux kernel and at the moment
where our Scripted Image Object would be accessed for its bash script,
a "no work to do" message would appear and the workstation would reboot.

- The img "View Hardware Information" would show different information
than what we had programmed in our rules. If we attempted to match the
information displayed in the "img" application, our rules would no longer
function. We learned that we should not attempt to match the information
and leave our rules alone.

We were dead in the water.

Examing our ZIMGDBG.LOG revealed numerous RAD -603 and RMAD -649 error
messages. Network packet traces revealed that the scripted image object
was indeed being accessed, but its contents were not being transmitted.
No matter what we did (create new server policy package, create new image
object, reinstall, reinstall fresh, etc.) we could not get ZDM 7.0 SP1
Imaging to work with PXE and Imaging Rules.

We worked with Novell engineers who in-turn engaged developers. It was
apparent to me that there had not been a whole lot of testing performed
of ZENworks Imaging on clustered systems running NetWare.

My advice to anyone running Imaging on a clustered system should hold off
on applying ZDM 7.0 SP1 until Novell fixes all of the defects (and there
are many).

UPDATE: The problem with the login dialogue box was related to DNS. We
don't have DNS running on our isolated network because it isn't needed.
Apparently the installer for ZDM 7.0 SP1 *MUST* have access to either a
DNS entry or HOST FILES entry for all nodes on a cluster. If it can't
resolve the name, the bogus login dialogue box appears. Apparently the
Novell developer in charge of testing on clusters failed to log a defect
for this problem which is why none of the front-line engineers and
technical support knew nothing of the issue.