[c-nsp] Redundant default route, without round-robin routing

Oliver Boehmer (oboehmer) oboehmer at cisco.com
Fri Sep 28 07:54:46 EDT 2007


Tim Franklin <mailto:tim at pelican.org> wrote on Friday, September 28,
2007 1:23 PM:

> On Fri, September 28, 2007 12:17 pm, Oliver Boehmer (oboehmer) wrote:
> 
>>> In that case, I'd be inclined to iBGP-mesh all the boxes, inject
>>> just a default into BGP from the two border routers, and let all
>>> the other routers pick their 'best' route of the two.  All other
>>> things being equal, this should be the border router that can be
>>> reached with the lowest IGP cost, or if they're all on a shared
>>> medium (I think this was the case from your original post), first
>>> route received. 
>> 
>> Even if you reach both next-hops via the same outgoing interface, IOS
>> will load-share between both of them. CEF will see two adjacencies,
>> no matter if they share the same interface.
> 
> You're confusing me now, Oli - BGP is still *not* multipath by
> default, right?  So one route makes it to the routing table, one
next-hop, one
> adjacency?

Oh, ok. I was assuming that multipath was enabled, so we'd have both
routes showing up the routing table. 
If it is not enabled, lowest router-id will win in this case (router-id
comparison will only be skipped if both paths were external).

	oli


More information about the cisco-nsp mailing list