MS NLB, multicast and routing

[Rigger1]

Hi all,

I've a question about Microsoft NLB, multicast (IGMP) and routing.

The NLB is configured to work in multicast (IGMP) mode, the switch (5400zl) has 'ip arp-mcast-replies' and 'ip igmp' (on all VLANS) enabled. If a client from the same subnet as the NLB send data (e.g. a simple ping) everything works as expected.

In addition, the switch as routing enabled and is the gateway for 3 more VLANS. If a client from on of these VLANS send data to the NLB, the switch floods it to all ports in (at least) the NLB VLAN.

So, can someone explain, why IGMP can prevent flooding within the same VLAN but not from other VLANs? It seems that I didn't understood that.

Kind regards,
Rigger1

[Vince-Whirlwind]

I think the switch assumes that other queriers could be anywhere, and so it floods it out.

Presumably, your switch is the "querier" by default, or maybe you have it off? Maybe you could try configuring it as querier in all vlans using vlan X ip igmp querier?

How about you try igmp filter-unknown-multicast?

Another thing worth knowing is that IGMP doesn't filter *all* multicast addresses. It doesn't filtger any address that as .0.0. or .128.0. in its middle octets.

Have you enabled ip multicast routing and router pim globally, as well as vlan X ip pim for each VLAN? 

I had a network of 5400s and 3500s where I configured multicast to get rid of all the flooded multicast traffic that was coming and going, and it worked really well for me. I don't recall the details, but I do remember spending a little time on it.

[Rigger1]

You have good suggestions. I'll try them out. But because some other projects, I think not before the end of the week.

I'll report my results :)

[Vince-Whirlwind]

Please let us know how you get on.

Configuring IGMP and PIM is a very over-looked task.