VMware Cloud Community
ForceV
Contributor
Contributor

Lost connections to multicast groups (igmp)

We're currently using multiple ESX servers in our production environment and are looking into implementing one or more in our product test environment. For this purpose I have installed ESX 3.5 update 2 (demo license) on a brand new 24 core INTEL machine. Our goal is to determine if we can replace multiple MPU's (Main Processing Units) with virtual MPU's. These MPU's communicate using multicast. That's working fine on the physical MPU's but not on the virtual MPU's i've created... One ESX server is connected to a switch (4 NIC's to 4 ports on the switch). The switch is from HP. 8 virtual MPU's are running RedHat EL 4u3. Multiple physical nodes (also connected to this LAN) are also running RHEL4u3. Our software relies on many MULTICAST groups for communication. It doesn't appear to be working 100%: not all multicasts seem to end up everywhere. When testing I noticed that not all VM's reply to the switches IGMP query . The switch ultimately excludes a port for the multicast group. What are your recommendations for switch, esx & vm settings for multicast communication? Is there a best practice doc? Known problems?

0 Kudos
2 Replies
DSTAVERT
Immortal
Immortal

Does your physical switch/switches impose any restrictions. Mac filtering etc.

-- David -- VMware Communities Moderator
0 Kudos
ForceV
Contributor
Contributor

No.

If we configure a firewall on the VM to reject the reports from the other VM's on the same ESX than it works. The problem is that the VM doesn't send a report when it sees a report from a neighbour VM. This is normal for IGMP but gives a problem with ESX software (VM switch).

0 Kudos