From: Schulz, Dave (DSchulz@dpsciences.com)
Date: Wed Jul 13 2005 - 01:23:10 GMT-3
Excellent! Thanks for the explanation.
Dave
-----Original Message-----
From: Paresh Khatri
To: Schulz, Dave; nobody@groupstudy.com; san ; lab
Sent: 7/12/2005 7:46 PM
Subject: RE: Basic BGP question
Hi Dave,
I'm sorry if my post came out that way but I was not saying that both R1
and R3 were clients to R2, just R1.
The situation (as described by the original poster) should work just
fine i.e R2 is a RR with R1 as a client. R3 is a non-client peer of R2.
You do not need to make R3 a client of R2. This is how it would work:
Here's the full scenario:
- R1 has one normal IBGP session to R2
- R2 has two IBGP sessions:
- one to R1 (as a route-reflector-client)
- one to R3 (as a normal IBGP peer)
- R3 has one normal IBGP session to R2
Route advertisement works as follows:
- any external routes received by R1 will be advertised to R2 via IBGP.
As per RR-functionality, R2 will advertise these routes to:
* all clients (in this case, there is only R1 from which we received
the update in the first place so it does not advertise it back to R1)
* all non-client peers - only R3 in this case
* all external peers - none in this case
- any external routes received by R3 will be advertised to R2 via IBGP.
As per RR-functionality, R2 will:
* reflect these routes to all clients (in this case, there is only
R1)
* all external peers - none in this case
- any routes originated by R2 will be advertised to all its peers - both
R1 and R3
So you have achieved what you wanted - complete dissemination of the
routing information.
HTH,
Paresh.
-----Original Message-----
From: Schulz, Dave [mailto:DSchulz@dpsciences.com]
Sent: Tuesday, 12 July 2005 10:51 PM
To: Paresh Khatri; nobody@groupstudy.com; san ; lab
Subject: RE: Basic BGP question
This brings up an interesting situation. Yes, you are correct...if you
have both R1 and R3 as clients to R2, then everyone is happy. However,
the original question had R3 as a non-client to R2....well, considering
the need to have a full-mesh (not using RR at this router)....I would
say that we need to have a connection set up to both R2 AND R1. What do
you think?
Dave
-----Original Message-----
From: nobody@groupstudy.com
To: san; lab
Sent: 7/10/2005 11:23 PM
Subject: RE: Basic BGP question
Hi,
Question 1: Yes, it will work. When using RRs, the requirement is to
have a full-mesh between all RRs and non-client IBGP peers, which you
have got.
Question 2: Within confederations, session establishment works just as
it does between actual EBGP peers. Therefore, if you are configuring a
session between peers in two different sub-ASs, you will need to use
ebgp-multihop if you are peering to non-directly-connected addresses
(e.g. loopbacks).
HTH,
Paresh.
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com
<mailto:nobody@groupstudy.com> ]On Behalf Of
san
Sent: Monday, 11 July 2005 01:09 PM
To: lab
Subject: Basic BGP question
Hi all,
Below basic BGP questions poped up my mind. I am sure most of you
know.
Question 1:
-----------------
R6----ebgp----R1 -------R2-------R3-----ebgp---R7
R1, R2, and R3 are IBGP.
R5 & R7 are external peers
Quote:
Routes from Non-client (R3) is advertised to clients only.
Routes from Clients is advertised to non-clients + other clients.
Assuming the quote is correct, Can i define just R1 as client to R2
& leave R3 as non-client to R2. Will everything work ok ? or Is
there a problem expected ?
Question 2:
-----------------
Within Confederation, according to previous observations & notes (IE
lab 8) i have written down "ebgp-multihop is not needed within
private peers of same confederation". I recently saw a
contradicating statement in one of the tech tips... that Can you
answer which one is correct ?
Thanks in Advance
/SAN
This archive was generated by hypermail 2.1.4 : Sun Sep 04 2005 - 17:00:29 GMT-3