Parallel EBGP sessions and RFC1771 collision detection.

From: Alex Bligh (amb@gxn.net)
Date: Thu Nov 25 1999 - 13:47:18 EST


This should be an easy question:

Can I safely maintain a two EBGP between two routers (A and B) if
I use different interfaces on each router? This appears to work in
the lab, though the RFC (quoted below) implies it might be unsafe. What
I am worrying about is one connection flapping, and in the process
of the flap, closing the other connection. My reading of this
implies there are circumstances where this can happen.

In my view, with EBGP in a non-multihop environment, one should be
safe using the link layer source and destination as additional
discriminators as to whether the connection should be closed,
as in this instance, it is, contrary to the RFC, not "clear" that
additional connections should be closed, as they may be set up
deliberately to maintain peering over multiple fabrics for resilience
(for instance). Does anyone know if this is the actual Cisco
implementation? (which is all I'm interested in right now)

IOS = 11.1.29CC1, 75xx.

Alex Bligh
GX Networks (formerly Xara Networks)

RFC1771 section 6.8 states:

" ... two parallel connections between this pair of speakers might
   well be formed. We refer to this situation as
   connection collision. Clearly, one of these connections must be
   closed.

   Based on the value of the BGP Identifier a convention is established
   for detecting which BGP connection is to be preserved when a
   collision does occur. The convention is to compare the BGP
   Identifiers of the peers involved in the collision and to retain only
   the connection initiated by the BGP speaker with the higher-valued
   BGP Identifier.

   ...

   A connection collision with an existing BGP connection that is in
   Established states causes unconditional closing of the newly
   created connection."

[ENDS]



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:07 EDT