cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
5537
Views
0
Helpful
2
Replies
Highlighted
Beginner

BGP Multipath with Route Reflectors

I have come accross an interesting affect of iBGP router reflection within a single AS. We have a number of multilayer switches accross a number of different PoPs. All running iBGP sessions back to our Route Reflectors in the Core. All routers in our network are configured for iBGP multi-path 2.

We have a customer which is dual connected within one of our PoPs. They have two routers, each connected to different access switches. The routers are assigned a /30 for their outside interface, to connect to us, and we then have static routes pointing a /27 customer allocation to each /30. Each access switch redistributes these statics into iBGP.

SW1 -> Router 1 next-hop for /27

SW2 -> Router 2 hext-hop for /27

When logged onto either of the Router Reflectors I can see both routes for the /27 in the BGP table and the routing table. The route reflector has installed bother customer routers as the next-hop address for the /27 and is load-sharing to them both.

RR1 -> /27 via SW1 and SW2

We recently had customer B, who connects at a different PoP, start talking to customer A. I have notices that the routers in other PoPs only have one router to customer A's /27. I have checked the BGP table and only recieve one route from the Route Reflectors.

SW3 -> /27 via SW1

Reading the details on RR I can see that this is stated as standard behaviour:

  • Routes received from a route-reflector client and selected as best routes are propagated to all BGP peers (internal and external).

I want to know if there is some IOS knob I can tweak to get the RR to advertise both BGP paths to this /27?

1 ACCEPTED SOLUTION

Accepted Solutions
Highlighted
Hall of Fame Master

Hello James,

this is a known issue with RR servers only within a L3 VPN you can use as workaround two different route distinguisher to have both VPNv4 routes propagated but if routes are in address family ipv4 unicast there is no easy workaround

this is written in docs about BGP multpath

see

Restrictions

Route Reflector Limitation

With multiple iBGP paths installed in a routing table, a route reflector will advertise only one of the paths (one next hop).

http://www.cisco.com/en/US/docs/ios/12_2t/12_2t2/feature/guide/ftbgpls.html

Hope to help

Giuseppe

View solution in original post

2 REPLIES 2
Highlighted
Hall of Fame Master

Hello James,

this is a known issue with RR servers only within a L3 VPN you can use as workaround two different route distinguisher to have both VPNv4 routes propagated but if routes are in address family ipv4 unicast there is no easy workaround

this is written in docs about BGP multpath

see

Restrictions

Route Reflector Limitation

With multiple iBGP paths installed in a routing table, a route reflector will advertise only one of the paths (one next hop).

http://www.cisco.com/en/US/docs/ios/12_2t/12_2t2/feature/guide/ftbgpls.html

Hope to help

Giuseppe

View solution in original post

Highlighted

I though that might be the case.

Tnx