Multicast issue?

[Supernn]

I have a weird one:
1. CCMS server in VLAN a
2. CCMA server in VLAN a
3. Citrix servers in VLAN a
4. Admin PC in VLAN a running WinXP Sp2, no fw
5. Citrix agents in VLAN b
6. Client PC's in other VLANS (e.g. VLAN c - i)

The situation is as follows:
CCMS sends multicast datastream to CCMA, Clients do receive multicast data correctly in VLAN c-i.
CCMA also sends data correctly to itself, and to some other W2K3 server in VLAN a but the only servers which works also have an ELAN connection. Other servers in VLAN a do not have an ELAN connection nor does mulitcast work.
The Admin PC cannot receive multicast data, if you move the admin PC to e.g. VLAN c it works fine.

The issue mainly lies with the Citrix servers, on which the software is deployed as mentioned in the NTP, but the datastream remains Unicast when starting the application on a Citrix deployed PC in VLAN b.

We do have enabled multicast throughout all VLANs, all BayStacks have been setup and tested several times....

Does anyone have an idea?

Supernn

 

[spanton]

What the Citrix server setup according to the CCMA NTP, Book 2 page 1754.

 

[Supernn]

Citrix setup as mentioned in NTP 297-2183-926_9.08 as of page 2099.

All works except for the multicast, but not only on Citrix, also on several other servers in VLAN a (of course not all).

 

[compmajor]

Are you saying the multicast only works in the ELAN?

 

[Supernn]

Not only ELan, also CLan but not in all Vlans....
The server VLan seems to have the problem, the Vlans used for the agents do not have an issue at all.

Believe me, if it was just to enable multicast in a VLan....

 

[dulfo666]

Check the binding orders of the network cards on your server. It sounds like the e-lan is binding first. Also check the metric for the multicast. Ensure that the c-lan has a lower metric than the e-lan.

I find the use of the multicast a real pain. Still, not too many network managers understand it and you only get blank looks.

 

[Supernn]

We've figured it out...
The Citrix servers were all connected to the same stack as the SWC. The switches are not multicast capable, so all traffic should be routed via the Passport first. This was solved by splitting up the network ports over both stacks, adjusting the teaming and adjusting the correct following order of the bindings.