[c-nsp] Filtering prefixes in mpBGP

[Oliver Boehmer (oboehmer)]

Pshem Kowalczyk <> wrote on Friday, October 19, 2007 1:39 AM:

> Hi All,
> 
> We have a 6503 working as a PE. One of the vrf will contain large
> number of prefixes (the Internet table) that I would like to prevent
> from spilling into the mpls cloud (all PEs that require full internet
> table peer directly from the vrf with each other). All I want is a few
> routes and a default. So this is what we set up:
> 
[...]
> 
> So in theory only two types of prefixes get marked on export from the
> vrf (internet and the loopbacks) with the route-target, and then the
> route-map towards the route server should filter all that doesn't have
> the right route target.
> But it looks like I missed something.

config looks ok at first glance. What exactly isn't working? Can you do
a "show ip bgp vpnv4 vrf Internet <prefix>" and verify the RT extcomm
set by the export-map?
Another option for you is to use inbound filtering on the RR using the
"bgp rr-group". 

What I don't understand is the "all PEs that require full internet table
peer directly from the vrf with each other". How is this done? What's
wrong about using the automatic RT filter on the PEs? I.e. all PEs which
don't import the "Internet" RTs will drop the updates anyway.. I
acknowledge that the RRs will still need to carry the prefixes..

	oli