All things related to multicast

[Tsegreda Beyene <>]

At 10:23 AM 10/31/2002 +1100, John Barlow wrote:

> We've had this on our Cisco OSR7600.  In the text "Interdomain Multicast
> Routing" (Addison-Wesley) on page 127 the authors describe the Cisco RPF
> check showing the behaviour you and I have seen:
>
> === START quote ===
> In Cisco's operating system, IOS, each routing protocol has its own
> routing table. An RPF check can use any routing protocol's routing
> table. When doing an RPF check for source address, S, the routing tables
> are scanned in order of protocol preference. The route used for the RPF
> check is the one from the most prefereed protocol that provides a route
> that encompasses S. It is important to note this route might not be the
> one with the longest mask.
> === END quote ===
>
> Like you, this was not the behaviour I expected, and really makes a
> mockery of the theoretical ability to manipulate the multicast traffic
> flow separately from unicast traffic.  Like you, my OSPF distributed
> default route was taking a different path from where I wanted the
> multicast traffic to flow in from.
>
>
> Cisco reps on this list - any chance the "longest mask" match might be
> implemented ?

John,

I have looked at this and There is actually a way  to enable longest-match.
Cisco uses distance first and then longest match per protocol table. 
However, given that, static is first, then BGP MRIB, DVMRP, and the various 
unicast routing protocols via distance. Consider the effect of using just 
longest match when one needs to set policy via MBGP. the unicast IGP 
protocol _may_ have a longest match and circumvent the MBGP policy.

Thanks.
Tsege


>   Or to change the RPF check to use the MBGP table first
> preference (which sounds good to me !)
>
> John Barlow
> GrangeNet (research network, Australia)
>
>
>
> > -----Original Message-----
> > From: 
>
> > 
> [mailto:]
>  On Behalf Of Roy D. Hockett
> > Sent: Saturday, 26 October 2002 9:23 AM
> > To: 
>
> > Subject: cisco multicast bug?
> >
> >
> > We have severial cisco 6513 w/ MSFC2, running hybrid mode OS, and are
> > seeing
> > a situation where we have MBGP routes for a multicast source, however,
> > IOS is using unicast default (via OSPF) for the rpf check. We can get
> > things
> > working with static mroutes, but obviously this is not ideal.  It was
> > suggested that this is a bug that was fixed in IOS for the 7500s, but
> > hasn't made it's way to IOs for MSFCs.  Does this sound familiar to
> > anyone?
> >
> > Thanks,
> > -Roy
> > --
> >
> > /********************************************************************/
> > /* Roy Hockett                     *   Telephone: (734) 763-7325    */
> > /* Network Engineer,               *         FAX: (734) 615-1727    */
> > /* ITCom,                          *    Internet: 
>
>   */
> > /* University of Michigan          *                                */
> > /********************************************************************/
> >