I'm not sure though... doesn't he want what the external peers sent to his border routers, not just what the border routers decided were the best routes?
Dan Farrell Applied Innovations [EMAIL PROTECTED] > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of > Eric Stockwell > Sent: Wednesday, September 27, 2006 4:19 PM > To: Tom Beard > Cc: misc@openbsd.org > Subject: Re: bgpd best external route > > Sounds like the behavior you are looking for is route reflection. > > Eric > > > > Tom Beard wrote: > > Henning Brauer wrote: > > > >> i honestly don't understand your problem ;( > >> > > I get told that a lot ;) > > > > Our two border routers (I'll call them B1 & B2) both have full views > > made up of various transit & peering connections. They have iBGP > > peerings with each other and also with both of the access routers (I'll > > call them A1 & A2). Under normal circumstances the access routers see > > ~180,000 prefixes from B1 and ~12,000 prefixes from B2. If for some > > reason B1 loses external connectivity, there is about a 2 minute time > > frame where A1 & A2 only have partial connectivity as B2 loses the > > routes from B1 and then starts advertising more of it's own external > routes. > > > > JunOS has an option that allow you to tell B1 & B2 to advertise a full > > table of routes to all iBGP peers so in the example of B2, it might have > > selected routes via B1 as active, however it will still advertise a full > > table of it's own best external routes. This means that should B1 lose > > connectivity, A1 and A2 already have a full route view from B2 and don't > > need to wait to it to re-converge. > > > > I'm not convinced that made much more sense. Perhaps I'm making the > > whole issue overly complicated? > > > > Tom