BGP Confederation config problem...

Rich Sena ras at poppa.clubrich.tiac.net
Sat Mar 7 11:27:40 UTC 1998


Hey Andy if it is a Confederation the BGP is treated like IBGP even if it
is EBGP so you can set a MED value to favor a specific route, you can also 
manipulate the AS_PATH attribute - appending the AS number 2 or more times
to the announcement for the least favored route applied to the appropriate
interface.

Since it's all considerred IBGP simply set the MED to 50 on the least
favored and 20 on the favored - or what ever incriments you deem
appropriate.

You may have to set "bgp always-compare-med" for it to be considered on
multiple AS numbers within a confederation though.

On Fri, 6 Mar 1998, Andy McConnell wrote:

> At each member AS, either IGRP or OSPF.  They do not interact.  By
> design... two of the AS's are across the Pacific from the other two :-)
> 
> -Andy
> 
> --
> Andy McConnell       $B%"%s%G%#(B $B%^%C%3%M%k(B
> Network Architect, NTT Multimedia Communications Laboratories
> 
> Slogan of 105.9, the classic rock radio station in Chicago: "Of
> all the radio stations in Chicago ... we're one of them."
> 
> On Fri, 6 Mar 1998, Blake Willis wrote:
> 
> blakew> Andy,
> blakew> 
> blakew> 	What IGP are you running on this network?
> blakew> 
> blakew> 	-Blake Willis
> blakew> 	CAIS Engineering
> blakew> 
> blakew> ---------------------------------------------------------------------------
> blakew> Blake Willis           					703-448-4470x483
> blakew> Network Engineer, New Customers				blakew at cais.net
> blakew> CAIS Internet, a CGX Communications Company
> blakew> ---------------------------------------------------------------------------
> blakew> 
> blakew> On Fri, 6 Mar 1998, Andy McConnell wrote:
> blakew> 
> blakew> > I have a problem I cannot solve.  If anyone is running confederations like
> blakew> > mine, I would appreciate some help :-)
> blakew> > 
> blakew> > We have a 4-memberAS confederation, each with two IBGP peers.  The
> blakew> > arrangement looks sort of like an octagon.  the "r" is AS3 is an internal 
> blakew> > hop, not a BGP peer.  
> blakew> > 
> blakew> >    _______     ______
> blakew> >   |      R-----R---r |
> blakew> >   |     / |   |    | |
> blakew> >   |AS2 R  |   |AS3 R |
> blakew> >    ----|--     ----|-
> blakew> >        |           |
> blakew> >    ____|__     ____|_
> blakew> >   |AS4 R  |   |AS1 R |
> blakew> >   |     \ |   |   /  |
> blakew> >   |      R-------R   |
> blakew> >    -------     ------
> blakew> > 
> blakew> > The problem is this:  How do you get BGP to choose the shortest "AS PATH",
> blakew> > since internal AS paths are ignored in selecting BGP routes?  Right now,
> blakew> > to top router in AS4 will always choose a route through (2 3 1) instead of
> blakew> > (1), because it prefers "external" routes (even external confederation
> blakew> > routes) over internal routes.
> blakew> > 
> blakew> > So, when given a choice, the router on the distant side of the AS will
> blakew> > ALWAYS prefer the three-AS-hop path, because it is external.  Is there a
> blakew> > way around this?!?
> blakew> > 
> blakew> > Avi Freedman suggested using a +1 metric when leaving each member-AS. 
> blakew> > (Thanks Avi!) But it doesn't seem to help - perhaps I didn't do it right. 
> blakew> > In fact, it doesn't look like the metrics are adjusted more than 1.  for
> blakew> > example, from AS2, the lower router sees every route outside of AS2 as
> blakew> > having only metric 1!  The top router in AS3 does not add 1 to the metrics
> blakew> > it readvertises to AS2.
> blakew> > 
> blakew> > I've been stewing over this problem for some time... I believe there is
> blakew> > some clue that I've missed.  Any help would be greatly appreciated!
> blakew> > 
> blakew> > -Andy
> blakew> > 
> blakew> > 
> blakew> > --
> blakew> > Andy McConnell       $B%"%s%G%#(B $B%^%C%3%M%k(B
> blakew> > Network Architect, NTT Multimedia Communications Laboratories
> blakew> > 
> blakew> > Writing about music is like dancing about architecture.
> blakew> > 
> blakew> 
> 

--
I am nothing if not net-Q! - ras at poppa.clubrich.tiac.net




More information about the NANOG mailing list