[c-nsp] partial prefix load in ibgp

[Mark Tinka]

On Wednesday, October 26, 2011 04:06:31 PM Nikolay Shopik 
wrote:

> I've got two borders connected via ibgp, both receving
> full-view via uplinks, and I notice one of borders
> receiving about 190K prefixes via ibgp. But I'm clearly
> sure there no filters at all. Configuration between
> borders are pretty simple and consist just two lines.
> 
> neighbor x.x.x.x remote-as 65536
> neighbor x.x.x.x next-hop-self

Just curious - are your border routers terminating customers 
directly? If not, and your customers are connected to edge 
routers lower in the topology stack, would you be happy with 
not allowing the border routers to exchange the routes they 
learn from their respective upstreams?

Rather, you simply send the upstream routes down to the edge 
routers to which your customers connect (via a route 
reflector, if you have one), and then let those edge routers 
choose which border router is the best path to which 
Internet destination?

This is what we do, and there's a certain joy to be had when 
upstream routes are not expressly learned among border 
routers.

Of course, if your border routers are also doubling as edge 
routers that your customers connect to, you have no choice 
but to do what you're doing now.

> Only difference is one border have configuration via
> address-family ipv4 while other doesn't have such
> sub-command, everything under router bgp. Any ideas
> where should I look?

Doesn't matter - contextual representation is local to the 
router. The BGP protocol still behaves the same way 
regardless of what CLI you choose to deploy.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20111026/a018381d/attachment.pgp>