[c-nsp] All multicast punting to CPU on 6500

[Phil Mayers]

On 17/12/12 11:38, Gert Doering wrote:
> Hi,
>
> On Mon, Dec 17, 2012 at 10:58:12AM +0000, Robert Williams wrote:
>> (PS. For the NLB fans, you'll be pleased to know the cluster is now going for good I'm told)
>
> I know I'm going to get stoned for that, but hey, it's Monday anyway...
>
> I actually *like* NLB.  It's a brilliant way to twist IP/ARP into
> "create resiliency and load distribution across a few servers, without
> having to buy and set up extra boxes to do resilient load balancing".

NLB has two modes of course - unicast and multicast. The unicast is 
completely uncontroversial - float an IP between two hosts and use 
grat-arp to move it, or float an IP & virtual MAC. Nothing to see, really.

The multicast is, I agree, a creative use of the various protocols. 
Sadly it's one of those things that probably looked good on paper and in 
the lab with hubs, but ended up only working on a subset of switches - 
it's not just Cisco gear it breaks.

I have a suspicion that re-using the IPv4/IPv6 multicast MAC OUI was 
part of the problem - that is special-cased in a lot of kit,