[Xorp-users] Questions on multicast routing

[Ben Greear]

On 07/01/2009 11:22 AM, Williams, Mark wrote:
> Well, looking at the diagram and details in the email, it looks like br1 on R1 was the rendezvous point for the multicast stream, and since you had a relatively high value of 1024000 (81 kbps with default interval value of 100) for switch-to-spt-threshold, a stream of only 64k would not trigger a switch to the shortest path.
>
> In any case, the register-stop message should have been sent from R1 to R2, but that didn't happen either. Was your fix related to this?

As soon as you remove an interface from pim, the task list would hang, and no further tasks
would be processed.  This included sending most (all?) packets to fea.

Once I fixed this I was able to add/remove interfaces and have it all keep working fine.

I think I'll make that switch-to-stp threshold smaller anyway though...seems a lot cleaner to
put even moderately busy connections in the kernel.

Thanks,
Ben


>
> -Mark Williams
>
> -----Original Message-----
> From: xorp-users-bounces at xorp.org [mailto:xorp-users-bounces at xorp.org] On Behalf Of Ben Greear
> Sent: Wednesday, July 01, 2009 11:16
> To: xorp-users at xorp.org
> Subject: Re: [Xorp-users] Questions on multicast routing
>
> On 06/29/2009 04:51 PM, Ben Greear wrote:
>> I am trying to figure out why multicast is taking the 'long' way through
>> the
>> network.
>>
>> My network diagram is attached. The xorp output below is from router 2
>> (the top one).
>>
>> I am generating mcast traffic from Node2 interface to Node3. I would
>> expect it to
>> go through the link between R2 and R3, but instead it goes R2 ->  R1 ->
>> R3. My traffic
>> flow is 64k, so I think it's using xorp user-space routing instead of
>> setting up
>> a kernel cache.
>>
>> I have hacked on multicast to support multiple routing tables, so it could
>> be my bug, but I'd still appreciate any pointers to what might be going
>> wrong
>> if anyone has any ideas!
>>
>> I'll be happy to send any other information, just ask.
>
> I think maybe I was confused on how multicast works..and I vaguely remember that this might
> have happened to me before :P
>
> Anyway, the real problem I was chasing was elsewhere, and the patch I sent to the mailing
> list yesterday fixed it.
>
> Thanks,
> Ben
>


-- 
Ben Greear <greearb at candelatech.com>
Candela Technologies Inc  http://www.candelatech.com