BGP path selection question

From: Jeff Lodwick (climberartist@xxxxxxxxxxx)
Date: Sun Nov 11 2001 - 04:29:30 GMT-3


   
Hi group,
I added a post with the subject "Weird BGP problem/question on BGP path
selection" that ended up being too long after I replied a couple times to
Eric on the list because of all the configs and debug I included. If you
would like the configs they are in the post with the subject "Weird BGP
problem/question on BGP path selection" or I can send them again if someone
needs them. My problem in a nutshell is with how BGP's path selection is
working on this lab setup I have. I have 6 routers that all are running
OSPF and no virtual-link problems so that I have simple routes to all
subnets. The only interfaces that aren't in OSPF are 2 loopback interfaces
off r4 (10.1.1.0 and 10.2.2.0) which I have redistributed into BGP
redistribute connected) and advertised through BGP. I have communities
sending med's to r4 so that
r4 selects certain routes to get to 2 different subnets off r1 and it is
working great. R1, r2 and r5 are all in AS 100 with 2 different
confederation AS's (65000 on r5 and 65001 on r1 and r2). R1 has a neighbor
to r2 and r5. R2 has a neighbor to r3 in AS 300 and r5 has
a neighbor to r6 in AS 600. R3 and r6 both have a neighbor to r4 in AS 400.
With this setup r1 picked the route through r2 as shown in the bgp table
showing 2 AS's through r2 (300 400) and 3 AS's through r5 (65000 600 400) I
also checked it with a traceroute to verify that was happening. To make it
so that the AS path was the same I set an
AS-path prepend on r4 pointing to r3 so that r1 sees both r2 and r5 with 3
AS's to pass through and should use the next determination in BGP's path
selection (origin then MED then closest IGP neighbor then BGP router ID).
The origin's are both incomplete since they are both redistributed into BGP
so the next path selection should be the MED. To test this I put a
route-map on r1 setting metrics to incoming routes from r2 and r5. I set it
up so that the subnet of 10.1.1.0 has a metric of 10 when coming from r2 and
any other routes coming from r2 are set with a metric of 20. I did the same
thing on r5 with the subnet of 10.2.2.0 having a metric of 10 and all other
routes a metric of 20. With this setup all of the routes on r1 learned
through BGP go through r2 then are withdrawn and then all routes learned
through BGP go through r5. I took off the second statement off of both
route
maps setting the metric of all other routes to 20 and it works fine. Routes
to 10.1.1.1 go through r2 and routes to 10.2.2.2 go through r5 but right
when I add those route maps back specifying all other traffic to a metric of
20 the same thing happens; all of the routes on r1 learned through BGP go
through r2 then are withdrawn and then all routes learned through BGP go
through r5.

Thanks in advance,
Jeff CCIE bound in 17 days



This archive was generated by hypermail 2.1.4 : Fri Jun 21 2002 - 06:45:12 GMT-3