All things related to multicast

[Matthew Davy <>]

If you're running MBGP on the 6509's, they should see MBGP routes for the 
commodity sources only from the I2 router and should prefer these (over the 
unicast BGP routes) for the RPF calculation.  

- Matt


On Mon, Jun 25, 2001 at 05:11:24PM -0400, Alan Crosswell wrote:
> I know many of you heard be talk about this at the Lincoln meeting or on
> the list here....
> 
> We were having this problem where multicast sessions from I2 that crossed
> our campus network would prune every 3 minutes and then come back 3 minutes
> later.  This appears to have been fixed.  How?  It's the unicast, stupid!
> 
> At least, fixing the unicast routing fixed the multicast problem.  Namely,
> we have two main exterior routers, one for I1 and one for I2 and a bunch
> of interior routers (8 or 9 6509's running Cosmos).  Our routing config
> was set up such that the interior routers did not have full Internet routes
> and used a default route to our I1 router which would then one-armed route
> traffic to our I2 router.  We knew this was dumb but didn't have our routing
> together enough to inject I2-leared routes into our interior routers.
> 
> So, in debugging the pruning problem, it become apparent to our Cisco
> TAC engineer after having me do a bunch of debugging that the SPT
> calculation was causing the tree to be rooted at the I1 router while
> the traffic was coming in to the I2 router.  The I1 router is not
> running any multicast protocols at all but it was on the reverse path
> to the source.  By doing a "ip pim spt-threshold infinity" we made the
> the SPT recalculation stop happening.  Then, since we just finished
> upgrading our interior 6509 routers to MSFC/2's with 512M we were able
> to turn on iBGP on them and, with a full routing table, the RPF was
> now correct for these I2 sources.  The errant pruning has stopped.
> 
> I wonder, though, with the MIX, if I were to get an I1 source would
> this work correctly?  The mcast would come in from I2 but the reverse
> path would be via I1. I can't seem to find an I1 multicast source that
> comes through from the MIX....
> 
> /a
> 
>