[c-nsp] eBGP multi-hop problem.

[oboehmer at cisco.com]

Rick,

connect your eBGP customer via ebgp-multihop to one of your
route-reflectors and announce only the customer's routes via iBGP from
there to the RSM. So you would have the BGP topology

 RR  ----ibgp--- RSM           customer
  |                               |
   --------------ebgp-------------

The RR must not modify the next-hop (i.e. no "next-hop-self" statement
on the ibgp connection towards the RSM), this way the RSM will see the
BGP prefixes from the RR pointing to the next-hop it will probably know
about as directly connected (otherwise you need a static route for the
next-hop on the RSM pointing to the customer) and will happily install
the BGP paths into its RIB/FIB.

	oli

Rick Ernst <> wrote on Friday, September 16, 2005 7:20 PM:

> I have an eBGP multi-hop customer connected behind a Cat5500/RSM.  I'm
> using multi-hop to a 7513 since the RSM can't handle full tables
> (128MB). 
> 
> The problem is the RSM doesn't have the route information for the
> network, so a routing loop is created between the core network and
> the RSM (the RSM defaults to loopback interfaces on the core routers
> via OSPF). 
> 
> My "get it working" fix was to static-route at a high pref to the RSM
> VLAN/interface, but that obviously breaks if the route is withdrawn. 
> I'm thinking about adding the RSM as a local-only BGP peer, and
> default everything else.  Is this the best way, or am I missing
> something more funamental?  How could this be configured if the RSM
> was some other device that couldn't handle BGP at all?
> 
> Thanks,
> 
> Rick
> 
> 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/