[c-nsp] BGP to OSPF redistribution

[Mark Lah]

Well on the BGP-side network, the router/switch that connects the OSPF
networks, you could create 2 separate OSPF processes.  1 process for the
remote network that will neighbor up across the L2VPN, and the other process
for the OSPF network that has BGP redistributing into it (the local network
from this devices perspective).  On this router/switch, then redistribute
the OSPF networks between the two processes (as noted earlier, be sure to
prevent loops with route-maps).

Now all the OSPF routes are seen as External (not necessarily ideal, but it
works), and you can then set the OSPF metric (cost) higher on the neighbor
adjacency(s) than taking routes learned from the BGP redistro.  You could
also do some summarization here too, which would prefer the more specific
route from BGP (may or may not be possible with your design).

-Mark


Date: Wed, 13 Jan 2010 16:52:02 -0500
> From: null zeroroute <nullzero.route at gmail.com>
> To: Vincent C Jones <v.jones at networkingunlimited.com>
> Cc: cisco-nsp at puck.nether.net, Mikael Abrahamsson <swmike at swm.pp.se>
> Subject: Re: [c-nsp] BGP to OSPF redistribution
> Message-ID:
>        <bd2721bb1001131352w474e7e8ew222e9182e1f8d458 at mail.gmail.com>
> Content-Type: text/plain; charset=ISO-8859-1
>
> Thanks for your suggestion.  We want to use OSPF because it will scale more
> easily in our network.  For example, if we ran BGP over the layer-2
> providers network, we would need (today) 25 neighbors at every site, every
> time a new site is added new neighbors need to be created everywhere, etc
> to
> keep the one hop away design.  Route-reflectors got too complicated.  It's
> also very helpful to have firewalls running OSPF when there are multiple
> egress points to extranet partner locations or the internet etc.
>