We have a multihomed ZCM 10.2 environment with 2 primary servers and 1 db server. The primary servers run SLES 10 and the DB is SQL2005 on W2K3.

All servers have two NICs. One NIC is connected to a private vlan dedicated to DB traffic. The other NIC is connected to a "public" vlan for commication between the servers and the managed devices. I have configured all the services to use the public vlan. My first primary server (named zcm01) has for the ip in the public vlan and for the private DB vlan.
In the following configurations files I have put in the "public" ip (

Unicast imaging works perfectly. Both manual and fully automated unicast imaging. I can create imaging bundles and assign them to devices...

But we have a problem with some parts of the multicast imaging.
I can setup a manual multicast session with a pxe client as master. The other pxe clients can connect and do imaging work. I can also setup a manual multicast session on one of the primary servers. This is done by running the command:
/opt/novell/zenworks/preboot/bin/novell-zmgmcast -mcast multitest -p /var/opt/novell/zenworks/content-repo/images/xp-base.zmg -c 5
The pxe clients can connect (img -s multitest) and do imaging. This is also OK and prooves that my network infrastructure is well configured for multicast.

But when I create a multicast image bundle in ZCC and assign it to devices it won't work. Well sometimes it works. That is 1 out of 20 times.
The pxe client boots up fine. But when the image job is about to start it just stalls for about 5 minutes (image session time out) and then bails out with a:
Imaging operation(s) assigned by
Elapsed time: 00:00
Failed to read from Preboot Server. See documentation

If I take a look at
it tells me that the multicast session is assigned to the address in the private DB vlan:
Thu Oct 15 13:05:37 2009 -- [RNT] XML is = <ZENworks-Imaging><Type>ZENworks-imaging-multicast-session</Type><SessionName>Multicast Image Set~4ee3d7e7485234b21ff86a2b768f54e3</SessionName><ServerIP></ServerIP></ZENworks-Imaging>

That explains why the bundle-assigned-multicast-job won't work. Any ideas why that happens? The few times that bundle-assigned-multicast actually does work, the novell-pbserv.log tells me that the multicast session was assigned to the correct (public) ipaddress. Is it a bug in Zenworks? Or do I need to configure anything else?

If I look at the primary server object in ZCC it correctly shows the two ipaddresses of the server (see attached file zcm01.jpg). Mostly the private ip is on top, but a few times it displays the public ip first. Is there a way to force the public ip address to be on top? It seems like they are randomly switched. Its a wild guess, but that could explain why multicast sometimes works.

Thanks in advance!