RE: [nsp] BGP loop detection

From: Philip Smith (pfs@cisco.com)
Date: Wed Jun 19 2002 - 01:48:33 EDT


At 17:15 18/06/2002 +0500, Ahmer Ghazi wrote:
>default-route will only be an effective solution if their customer is not
>getting BGP routes from any other provider.

The original question didn't mention that the subdivided network had other
connectivity. It was a simple question on how to provide connectivity
between two networks sharing the same ASN through an intermediate ISP.
Hence the simple answer. ;-)

> Otherwise, everything will point towards the BGP routes learned
> through the other provider(s), and the default route will not be used.

If there is other connectivity, the subdivided network points a static
route for the other /19 to their upstream (the transit ISP/ASN in question)
- yes, they still use BGP for everything, but the static route ensures that
they can reach the other portion of their network through the preferred
upstream.

The upstream/transit ISP in question doesn't have to do any magic at all.
As far as they are concerned they have two BGP customers, so they treat
them like any other BGP customer. The onus is on the subdivided network to
ensure that they have the necessary static to reach the other part of their
network.

BTW, RFC2270 describes the solution to this problem in good detail. I've
paraphrased that in my answer - the BGP tutorial slides I referenced has
the configuration examples.

hope this clarifies,

philip

--

>Ahmer Ghazi > > >-----Original Message----- >From: pankaj [mailto:pankaj@worldgatein.net] >Sent: Tuesday, June 18, 2002 3:32 PM >To: Zaheer Aziz >Cc: cisco-nsp@puck.nether.net >Subject: Re: [nsp] BGP loop detection > >More easier solution is as Nick and Philip said. > >Use static route for eachother's /19 pointing to your router. Or just >use default route pointing to ur router. >Thats enough for the situation u describe > >--pankaj > >----- Original Message ----- >From: "Zaheer Aziz" <zaziz@cisco.com> >To: "Nick Kraal" <nick@arc.net.my> >Cc: <cisco-nsp@puck.nether.net> >Sent: Monday, June 17, 2002 10:44 PM >Subject: Re: [nsp] BGP loop detection > > > > At 10:17 PM 6/17/2002 +0800, Nick Kraal wrote: > > > > Another option,if you continue to run EBGP with them, is to have >them > > configure > > > > neighbor EBGP-address allowas-in > > > > at their EBGP boxes peering with you. > > > > This will accept paths with local as in the AS_PATH. It is mainly >used in > > MPLS-VPN scenarios but can certainly be applicable here. > > > > Zaheer > > >We have a customer that has transit services from us. They have >split their > > >network into two /19 and run BGP with us at two different >locations. Each > > >/19 is actually for two seperate businesses; one a tier-2 ISP and >the other > > >VoIP provider. Two physically sperate networks are also running the >same ASN > > >and are to peer with each other via our network. There is no >internal > > >interconnectivity between them or any IGP running. > > > > > >The problem lies in that to reach each other they need to 'transit' >via our > > >network and according to BGP, one cannot announce back the client >prefixes > > >learnt from the other network and vice versa as both of them are >running the > > >same ASN. This is to prevent routing loops. > > > > > >In this case is there a method to overwrite this so that the first >/19 > > >network will be able to reach the second /19 network via ours? > > > > > >Thanks in advance. > > > > > >-nick/ > > > >



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:13:47 EDT