[c-nsp] How to tune multicast RPF checking?

[Félix Izquierdo]

Matti Saarinen wrote:

> We have a network setup where multicast is transported through PIX via
> GRE tunnel. Now, I have some difficulties in tuning the RPF checking
> of the multicast packets
> 
> The network looks roughly like this
> 
> 
> rtrA --- fw --- (small network: rtrB + 2 routers) --- rtrB (Cat6500)
>    \                                                         /
>     \--------------- GRE tunnel for multicast --------------/
> 
> 
> On rtrA there are static mroutes pointing to the tunnel, It consis of
> the prefixes that are behind the PIX. On the rtrB there is a mroute
> towards 0/0 that points to tunnel as well. The default route is
> generated by the PIX and advertised via OSPF which is the IGP used.
> The RP is rtrB which is a Cat6500 and there are six other routers as
> well.
> 
> Now, when a host in our internal network starts sending multicast
> traffic the traffic will be dropped when it reaches the RP due the RPF
> check. This due the fact that the default mroute overrides the more
> specific routes learned via OSPF because of the lower administrative
> distance. If I set a more specific static mroute pointing towards the
> source the RPF check succeeds and the multicast traffic flows
> correctly.
> 
> Setting up a static mroute for every network in which there might be a
> multicast source gets really soon annoying. Is there any way to make
> the RPF check work on the RP without the static mroutes? I know in
> JUNOS I could configure a special RIB for RPF but can I configure a
> similar one in IOS?
> 

This is that multicast address family of bgp does. Perhaps a solution is 
to run bgp ( only multicast family ) and redistribute ospf in multicast 
address family.

Cheers.

Félix