[tech] The BGP experiment...

Adrian Chadd adrian at ucc.gu.uwa.edu.au
Fri Feb 15 13:57:04 WST 2002


On Thu, Feb 14, 2002, Grahame Bowland wrote:

> [  UCS ] -> <ibgp> -> [ random servers wanting description
>                         of WAIX ]
> 
> Now, I can do an iBGP session fine between two of my boxen  
> using zebra. However, the routes learned over the multihop 
> session aren't distributed to the other zebras because 
> WAIX isn't directly reachable, so the routes are not 
> of interest..

Uh. The routes are of interest if you have ospf routes
providing next-hop info for all the next hops..

Here's a question - what does your zebra routing table look
like? Have you populated zebra with your default route and
such? Does zebra have a 'no sync' option like the cisco
routers do?

> So, what is the trick? Should I use a route-map to change 
> the next hop to something local? I'd rather avoid this as 
> it seems like a hack and something might actually send 
> the traffic to me, which won't work.

Define "send the traffic to you". As long as you actually
repeat the next-hop setting route map evilness on all the
boxes that you speak iBGP to, it'll be ok. You'll probably
want to do this anyway, since any next-hop info you manage
to propagate from WAIX (even without routemaps) will
be kinda useless.

2c,



Adrian



More information about the tech mailing list