Load balanced App HPSM looses intra app connectivity across nodes

[nathanw]

Hello,

I have a guest cluster running HP Service Manager of which

HP uses a broker (as I understand it) and multicast for its own application

Load balancing.

I have a 6 node (guest Cluster) when all the guests are on

the same Physical Host the application service works as expected, however when

one of the servers is migrated to another Host it will loose contact with the

farm, most annoying.

Now when this occurs Multicast – egress packets are no

longer getting to the migrated server but are being sent from the master node. The

communications for the management of this data is Multicast.

Are these any known issues with Multicasting packets? As my

network guys are basically saying it’s not them.

I am running ESXi 4.0 Update 2

DVS switches and have also tried standard virtual switches,

single and multiple Uplinks.

Ping functions correctly

I have tried accepting all security parameters – Promiscuous

Mode, Mac Address Changes, and Forged Transmits. +This is NOT Microsoft NLB and I believe that the app does not do Mac

Changes or Shared IP address across nodes+.

I have also tried turning off Notify Switches.

Has anyone out there run this app in a VM before?

Any suggestions where this could be going wrong.

Many Thanks

[Herschelle]

Have you seen this document? Pg 11.

http://www.google.com.au/url?sa=t&rct=j&q=%22HP+service+manager%22+vmware+vmotion&source=web&cd=1&ve...

Not a very helpful document and does not explain any technical details.

[iw123]

What is the multicast address you are sending to? Is it a true multicast address or a unicast address (as yoiu tend to see with MS NLB). It may be that you need a static arp entry for your multicast address set up.

[orthohin]

Is there any VLAN configuration?

Regards,
Milton