RE: vpls

From: Aaron <aaron1_at_gvtc.com>
Date: Mon, 19 Mar 2012 19:41:14 -0500

Thanks Ronnie. I appreciate the input. Do you mean I should disable split
horizon on the pw's coming in on the hub vpls vfi location in order to allow
pw to pw switching? I mean will I have to do that manually? Like somehow
disable split horizon per pw? if so lemme know commands if you know them..

 

Is it something like..?

 

Conf

L2vpn

BG..
BD..
(ac here)

Vfi.
neighbor 1.1.1.1 123
(disable SH somewhere in here?)

 

 

Also, if I wanted a dual hub spoke architecture, it would seem that I would
have each spoke site dual homed to those dual hubs I mentioned, and then
those dual pw's coming into a single spoke site would simply NOT forward
between those pw's by default. .with a single connection between the dual
hubs (with SH disabled on that hub to hub connection)..does that seem
do-able to you?

 

(I actually need to set this up tomorrow...I have (15) asr9k's and about
(10) asr901's that I need to get up and running over this vpls architecture
I'm discussing with you..this is the inband (quasi oob mgmt) architecture
that I'm talking about doing vpls with)

 

Thanks

Aaron

 

 

From: Ronnie Angello [mailto:ronnie.angello_at_gmail.com]
Sent: Monday, March 19, 2012 7:27 PM
To: Aaron
Cc: Cisco certification
Subject: Re: vpls

 

Full mesh is the most commonly deployed architecture, but you can also have
a hub and spoke or partial mesh topology. Split horizon isn't enabled in
these cases. It's disabled on the hub in a hub and spoke topology as it's
loop free by definition. You must run STP over the backbone with a partial
mesh.

Ronnie

On Mar 19, 2012 10:01 AM, "Aaron" <aaron1_at_gvtc.com> wrote:

As I understand it, pw's attached to a vpls group (I think known as a vfi)
within an ASR9K (or perhaps any vpls capable device) have a slit horizon
rule that disallows pw---to---pw switching (by default); something about
loop prevention. and thus because of this requires a full pw mesh between
vpls nodes. Is there a way around this or is that a hard fast rule? (ibgp
requires full mesh, but there's a way around it with rr's)

Aaron

Blogs and organic groups at http://www.ccie.net
Received on Mon Mar 19 2012 - 19:41:14 ART

This archive was generated by hypermail 2.2.0 : Sun Apr 01 2012 - 07:56:52 ART