From: Prashant Shukla (shukla_cisco@yahoo.co.in)
Date: Thu Mar 08 2007 - 03:05:03 ART
Hi,
As per my understanding, when creating iBGP peers ,the source address of the TCP connection is the IP of the outgoing interface, this is an acceptable peering requirement.
we use Loopback interface as the source of all BGP peers in case ( we generally do ) if we have redundant paths to reach the same peer, then we are not bound to a specific interface for the source IP if that specific outgoing interface goes down..
Ex.R1(config)# router bgp 2001
R1 (config-router)# neighbour a.b.c.d
This "neighbor" address is the source of the TCP connection that this router is expecting, so we give Loopback as its the most stable source IP address even if one of the links goes down and we have other redundant paths to reach the same neighbor.
Secondly for eBGP peering its mandatory to have eBGP neighbors directly attached, as the TTL value of the eBGP packet is 1. Thus IF we have non-directly attached eBGP neighbors then we need to issue the "eBGP multihop" command to increase the TTL value to more than 1.
the bottom line is, to be on the safer side, use the most "stable" interface as the source of iBGP connections, thus thats why "loopback" is used.
HTH
Shukla.
achievewoo@gmail.com wrote: Hello, GS
I really confuse when I try to choose neighbor address to build BGP neighbor relationship.
For example, Two redundance links exist between BGP peers, but, there is no any requirement to use loopback interface as neighbor address, Should I use it?
Is there any standard when choosing neighbor address; interface address or router ID? Is there difference between IBGP peer and EBGP when choosing neighbor address?
Thanks!
This archive was generated by hypermail 2.1.4 : Sun Apr 01 2007 - 06:35:50 ART