BGP labeled unicast (RFC3107) - Intra Area Traffic shouldn't touch ABR Router

[DOMJAHN DAVID]

Hi there,

setting up RFC3107 (BGP Labeled Unicast) we run into the problem, that intra area traffic is always handled by the Area Border Router (ABR) acting as the inline Route Reflector (RR1).

This is our setup:

        RR2

       |

PE1----P1-----ABR1/RR1--> core

       |

PE2----+

VPNv4 traffic sent by PE1 will not be directly forwarded to PE2 but will be sent to ABR1 which reflects

the traffic to PE2.

Configuration:

• ABR1 acts as the inline Route Reflector, being configured with the "next-hop-self" feature (needed for BGP labeled unicast functionality);
• BGP sessions are established between

1) PE1 and ABR1

2) PE2 and ABR2

We tried to introduce an additional Route Reflector RR2 (not using the next-hop-self feature) being attached to P1 and let this RR peer with

• PE1
• PE2
• ABR1 (which still acts as the inline RR for RFC3107)

but as RR2's routes may be overwritten by ABR1/RR1.

Question: Do you have any ideas how to forward intra area traffic directly (without taking the way to ABR1)?

Thank you very much!

Kind regards,

David

[Nagendra Kumar Nainar]

Hi David,

Can you try making RR2 as route reflector for PE1, PE2 and ABR1. From PE1 and PE2 remove the session to ABR1.

Set "next-hop-self all" on ABR1 towards RR2. This way, any prefix from CORE to AGG (via ABR1) will be rewritten with its own address. PE1 to PE2 will be advertised by RR2 without changing the NH.

-Nagendra

[DOMJAHN DAVID]

Hi Nagendra,

thanks a lot for the quick reply!

We already moved iBGP sessions from ABR1 to RR2 without success. Further investigation is needed - I'll keep you up to date.

Regards,

David