[j-nsp] Multicast

Caren Litvanyi litvanyi at grnoc.iu.edu
Fri Jul 15 18:09:22 EDT 2005


On Fri, 15 Jul 2005, Julian Cowley wrote:
> On 15 Jul 2005, at 01:03, Pablo Varela wrote:
> > On NANOG 27 there was a presentation called "A Methodology for
> > Troubleshooting Interdomain IP Multicast" that presents a methodology
> > that looks very throrough.
> > Unfortunately I don't have the RUL but it shouldn't be too hard to  
> > find.
> 
> Yes, I have to attest this presents a very straightforward way to  
> troubleshooting multicast.  We've used it here successfully to fix  
> our own multicast feed through Sprint.
> 
> http://www.nanog.org/mtg-0302/ppt/caron.pdf
> 
Hi,
I gave that talk way long ago; I'm glad some people have found
it useful.  We have been updating that talk a bit and presenting
it as part of the Internet2 multicast training workshops in the
higher ed community, though the material is essentially the same.
Here's a pointer to the latest version, though I still consider
it quite incomplete and a work always in need of revision (which
I never get to...)
http://multicast.internet2.edu/workshops/indiana/docs/internet2-multicast-workshop-june-2005-5-tshooting.ppt
If you just go to "workshops" you can usually click down to
the latest version.  Thanks in advance for any suggestions for
improvement.  Sorry for ppt only.
Just out of curiosity, does J2 have a route in inet.2 for both
the RP address on J1 and for the traffic generator?  Just to
make sure your rib groups are ok and connected interfaces got
in there.  (I probably should take my own advice and not shoot
in the dark, though...)
Thanks,
Caren
(Credits to Bill Nickless for the original paper that sparked
my work on the slides.)

> > Pablo
> >
> > On Fri, 15 Jul 2005 11:53:23 +0100, "Jee Kay" <jeekay at gmail.com> said:
> >
> >> I've got a really strange problem with multicast between a pair of
> >> Junipers. The setup is as follows:
> >>
> >> Traffic Generator -- Juniper1 -- Juniper2 -- Traffic Sink
> >>
> >> The two Junipers see each other as PIM neighbours happily. Juniper1
> >> seems a PIM Join entry for the Sink, and the output of show pim join
> >> extensive and show multicast route says it is correctly forwarding  
> >> the
> >> multicast traffic out towards Juniper2.
> >>
> >> Juniper2, in turn, has a PIM Join entry correctly showing that  
> >> traffic
> >> should be sent towards the sink, aswell as the appropriate IGMP Group
> >> entry. Show multicast route on this route says no route entries
> >> matched.
> >>
> >> As far as I can tell from a sniffer between J1 and J2, the multicast
> >> traffic is genuinely appearing on the link, although a filter on the
> >> ingress interface on J2 doesn't seem to pick it up.
> >>
> >> Anyone have any ideas how to go about troubleshooting exactly what's
> >> happening here? It looks like all the individual components are
> >> working correctly but there is some key part missing that is causing
> >> J2 to receive the multicast but not realise that, despite the IGMP
> >> state, it needs forwarding on.
> >>
> >> For the record, this is all PIM Sparse, J1 is the RP (set as local on
> >> J1, static on J2), I am using inet.2 for RPF checks (with the
> >> appropriate rib-groups to distribute OSPF into .0 and .2), have
> >> interface-routes set to distribute into both tables also. IGMP is
> >> enabled on all interfaces, version 2. Packet traces show that the  
> >> sink
> >> is correctly using IGMPv2. PIM is also enabled on all interfaces,  
> >> mode
> >> sparse. The RP address is a secondary IP on lo0 of J1. PIM and IGMP
> >> are disabled on fxp0.
> >>
> >> Thanks in advance,
> >> Ras
------------
Caren Litvanyi, network engineer
Global Research NOC at Indiana University
litvanyi at grnoc.iu.edu


More information about the juniper-nsp mailing list