All things related to multicast

[Tyrone Kelly <>]

I believe your fix would be resolved if your ISP sends partial routes + a 
default route.  Where the BGP next hop would then be the next-hop toward the 
rp/originator, which would satisfy rule condition #2 of the MSDP-rpf peer 
rules of rfc 3618. This is provided you aren't  doing next-hop self over your 
ebgp peering toward your provider?

Also you wrote: 

"show ip rpf" points to the default network when no prefix matches the 
source, with a "unicast" RPF type. However MSDP peer-RPF check fails: SA 
announcements are rejected when no prefix matches the RP/Originator, so the 
default network doesn't apply in this case, or so it seems.

You have to remember RPF- check for S,G are different from MSDP RPF-peer rule 
checks.  The difference is the RPF check used when forwarding data packets, 
which compares a packet's source address against the interface upon which the 
packet was received.  The MSDP rpf-peering rules compares the RP 's address 
from the SA message to the MSDP peer from which it was received. So one of 
the 5 peering rules of MSDP RPF-Peer rule checks allows for the RPF-Peer to 
be valid if the BGP next hop is the next-hop toward the RP/originator of the 
SA. 

-----Original Message-----
From: Zenon Mousmoulas 
[mailto:]
 
Sent: Friday, June 11, 2010 2:24 AM
To: 

Cc: wg-multicast
Subject: Re: inter-domain MSDP peering without BGP FIRT

On 10 Ιουν 2010, at 9:32 ΜΜ, Bill Owens wrote:

> On Thu, Jun 10, 2010 at 08:57:10PM +0300, Zenon Mousmoulas wrote:
>> However MSDP peer-RPF check fails: SA
>> announcements are rejected when no prefix matches the RP/Originator,
>> so the default network doesn't apply in this case, or so it seems.
>
> How is the default route created - static, or supplied by the eBGP  
> peer? I just tried a quick experiment with an internal router that  
> has a static default:

Quite similar to "ip route 0.0.0.0", the default route is installed  
with "ip default-network" (gateway of last resort) pointing to a  
network which is known by eBGP (announced from GRNET).

> c7609#sh config | incl 0.0.0.0
> ip route 0.0.0.0 0.0.0.0 199.109.35.1
> c7609#conf t
> Enter configuration commands, one per line.  End with CNTL/Z.
> c7609(config)#ip mroute 0.0.0.0 0.0.0.0 199.109.35.1
> c7609(config)#end
> c7609#sh ip rpf 2.3.4.5
> RPF information for ? (2.3.4.5)
>  RPF interface: GigabitEthernet1/1
>  RPF neighbor: nn-3550-ext.nysernet.net (199.109.35.1)
>  RPF route/mask: 0.0.0.0/0
>  RPF type: static
>  RPF recursion count: 0
>  Doing distance-preferred lookups across tables

RPF information for ? (2.3.4.5)
   RPF interface: <GRNET uplink>
   RPF neighbor: w.x.y.z
   RPF route/mask: 194.177.208.0/22
   RPF type: unicast (bgp <U AS>)
   RPF recursion count: 0
   Doing distance-preferred lookups across tables

> c7609#conf t
> Enter configuration commands, one per line.  End with CNTL/Z.
> c7609(config)#no ip mroute 0.0.0.0 0.0.0.0 199.109.35.1
> c7609(config)#end
> c7609#sh ip rpf 2.3.4.5
> RPF information for ? (2.3.4.5) failed, no route exists
>
> Seems to indicate that the static mroute works. It would be  
> incorrect in this case, since the unicast default route points to  
> our commercial ISP connection, and the R&E side is native BGP/mBGP.

So the difference is you have a static default route pointing to the  
IP next-hop, while they are pointing it to a network learnt via BGP  
from GRNET. However "ip default-network" does not have a multicast  
counterpart; perhaps they could use "ip route" and "ip mroute" instead.

However, the question is: does MSDP work for you in this case? Would  
it accept an SA where the originator would match this default route in  
an RPF lookup?

Thanks,
Z.