Tricky BGP into IGP Question
What is the preferred way to get transit customers routing information into your IGP so that BGP will announce the route to it's neighbors. 1===2a===2b====3 | 4 1 is the customer, announcing to 2a, who propogates 1's route to 2b and to 4. 2b will not announce the route to 3 as it doesn't have a route in it's routing table for 1's route. I was thinking about redistributing bgp into my igp using a route-map based on as-path filter, but that seems somewhat kludgy. What are you all doing to accomplish this or am I just missing something here? Thanks, --Tim ============================================= Timothy M. Wolfe CCNA, NSA Sr. Security Engineer tim@ignw.com InfoGroup Northwest 541.485.0957 x108 =============================================
On Tue, Jan 23, 2001 at 03:26:28PM -0800, Tim Wolfe wrote:
What is the preferred way to get transit customers routing information into your IGP so that BGP will announce the route to it's neighbors.
1===2a===2b====3 | 4
1 is the customer, announcing to 2a, who propogates 1's route to 2b and to 4. 2b will not announce the route to 3 as it doesn't have a route in it's routing table for 1's route. I was thinking about redistributing bgp into my igp using a route-map based on as-path filter, but that seems somewhat kludgy. What are you all doing to accomplish this or am I just missing something here?
Assuming each of those routers only speaks BGP to its direct neighbors currently, I'd say the most common solutions are route reflectors (and/or confederations) and fully-meshed BGP networks. The reason 3 doesn't see the route is because BGP doesn't propagate routes across multiple iBGP hops. If one of (2a, 2b, 3, 4) acts as a route reflector, and the others peer only with it, your route will get propagated to all of them. Alternately, if each of (2a, 2b, 3, 4) peers with each other directly, 1 will announce to 2a and 2a will propagate to all others. -c
Hello Tim, I'm assuming that "1," "3," and "4" are separate AS numbers with "2a" and "2b" in the same as providing transit ....... given that ..... The BCP for redistributing BGP into the IGP is just _NOT_ to do it. Redistributing BGP into the IGP is a very dangerous thing to do. Besides, the problem you described below has nothing to do with "1" routes being in the IGP. They should not be in the IGP in the first place. The problem you are describing sounds just like a recursive lookup failure on "1's" next-hop information. The subnet between 1 and 2a is not in the IGP. Hence, when 1's routes get to 2b, 2b will do a recursive look-up on 1's routes and have it fail (cause the next hop is not in the forward table). So one router 2b, check for a entry and pingability of subnet between 1 and 2a. The link between router 1 and 2a is an 'infrastructure' route, so you want that in your IGP. Barry
-----Original Message----- From: owner-nanog@merit.edu [mailto:owner-nanog@merit.edu]On Behalf Of Tim Wolfe Sent: Tuesday, January 23, 2001 3:26 PM To: 'nanog@merit.edu' Subject: Tricky BGP into IGP Question
What is the preferred way to get transit customers routing information into your IGP so that BGP will announce the route to it's neighbors.
1===2a===2b====3 | 4
1 is the customer, announcing to 2a, who propogates 1's route to 2b and to 4. 2b will not announce the route to 3 as it doesn't have a route in it's routing table for 1's route. I was thinking about redistributing bgp into my igp using a route-map based on as-path filter, but that seems somewhat kludgy. What are you all doing to accomplish this or am I just missing something here?
Thanks,
--Tim
============================================= Timothy M. Wolfe CCNA, NSA Sr. Security Engineer tim@ignw.com InfoGroup Northwest 541.485.0957 x108 =============================================
participants (3)
-
Barry Raveendran Greene -
Clayton Fiske -
Tim Wolfe