I'm having a problem on my cluster and I'm being told that the reason why is that my one of my machines is using V2 while the rest are using V3. I've used the same disks to do all the installs so I don't know how they could be different. I don't see an RPM for IGMP, so how do you upgrade it? Thanks!

In Linux IGMP support is in the kernel. Which version do you have?

Thanks for the response Mara... I'm using version 2.6.5-1.358. Should I upgrade that? I don't understand how all my nodes are running he same kernal version yet have different IGMP versions.

2.6.5 has IGMPv3 (there are other reasons to upgrade). The thing is why it doesn't work for you... Look into /proc/net/igmp. What does it say?

Mara,

Thanks for your help. When I do a 'cat /proc/net/igmp' I get the following:

###########################
Idx Device : Count Querier Group Users Timer Reporter
1 lo : 0 V3
550100E0 4 0:00000000 0
540100E0 8 0:00000000 0
010000E0 1 0:00000000 0
4 eth0 : 6 V2
550100E0 4 0:00000000 1
540100E0 8 0:00000000 1
470B02EF 1 0:00000000 1
010000E0 1 0:00000000 0
5 eth1 : 5 V2
550100E0 4 0:00000000 1
540100E0 8 0:00000000 1
010000E0 1 0:00000000 0
###########################

What do you think?

It's clear that it says it supports version 3 (V3 near lo). Why eth0 and eth1 use version 2? Good question. Probably it didn't receive proper replies just after boot and made a fallback to version 2. There's a method to force it to version 3 without reboot. Example:
/sbin/sysctl -w net.ipv4.conf.eth0.force_igmp_version=3
/proc/net/igmp should then show V3 for eth0. Use similar command for eth1 and check if it works correctly (you may dump some traffic using tcpdump to see if version 3 is really used).

Thanks Mara... you Linux God!

I'm run that command, but my 'cat /proc/net/igmp' is returning the same output that I have listed before. It's interesting because I get no error when running the command. This is what I see.

##############################
[root@cluster1 root]# /sbin/sysctl -w net.ipv4.conf.eth0.force_igmp_version=3
net.ipv4.conf.eth0.force_igmp_version = 3
[root@pleiades root]# cat /proc/net/igmp
Idx Device : Count Querier Group Users Timer Reporter
1 lo : 0 V3
550100E0 4 0:00000000 0
540100E0 8 0:00000000 0
010000E0 1 0:00000000 0
4 eth0 : 6 V2
550100E0 4 0:00000000 1
540100E0 8 0:00000000 1
470B02EF 1 0:00000000 1
010000E0 1 0:00000000 0
5 eth1 : 5 V2
550100E0 4 0:00000000 1
540100E0 8 0:00000000 1
010000E0 1 0:00000000 0

##############################

It may be because IGMP is in use. The switch may work after timeout (what can take some time). Do you have a chance to reboot the machine and try it just after?

Yup... Rebooting didn't work. Do you think that it could be because I need to upgrade my kernel?

Nope, I don't think that you need an upgrade. You have version 3 support (at least it's saying so). I'm wondering if you have experience with ethereal or any other sniffer? It may be a good idea to run one between the hosts and see what's comming in just before reboot (especially: see IGMP version used).

Could you recommoned a small sniffer? I could install snort but I would think there should be a lite version of that. Thank you so much for your help Mara. =)

Snort is not a sniffer (well..it can work as a sniffer, but that would be an overkill). I recommend ethereal (graphical). It also has console mode. But that's a nice tool for analyzing the trace file on your desktop machine. To get the traffic you can just use tcpdump. The command will be something like
tcpdump -i eth0 -w output.log
Use the right interface name for the one you want to sniff on. When you think it's done stop tcpdump (ctrl+c) and copy the file to your desktop machine when you can look into it.