From: Nico van Niekerk (nico@vanniekerk.co.za)
Date: Tue Nov 23 2004 - 19:49:54 GMT-3
Let my see if I understand your scenario...
A,B,C are iBGP peers?
D is non-BGP
A,B,C,D are all in the same OSPF area?
Why redistribute from BGP into OSPF on A,B and C?
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
> Joe Chang
> Sent: Wednesday, 24 November 2004 3:23 AM
> To: ccielab@groupstudy.com
> Subject: Re: Redistribution of BGP into the IGP for a transit network
>
>
> > The fact that you are a transit doesn't really have any
> > bearing on the route redistribution.
>
> The scenario where a transit AS might require restrictions on
> redistribution is this:
>
> Say all three are BGP routers in the same AS. All three routers
> also have EBGP peers with different provider systems:
>
> A-----B
> \ /
> --C--
>
> The IBGP peers are not directly physically connected.
> Synchronization is not to be disabled. EBGP routes must be
> redistributed into the IGP.
>
> Say redistribution is configured on each peer with no restrictions. e.g.:
>
> router ospf 1
> redistribute bgp 1 subnets
>
> Say there is a route 10.0.0.0 that C receives from its EBGP peer.
> This route would be redistributed into the IGP at peers A, B as
> well as C.
>
> There is an intermediate, non-BGP router D that needs to find a
> route to 10.0.0.0:
>
> A B
> \ /
> D
> |
> C
>
> D must use the IGP to determine in which direction 10.0.0.0 is.
> Instead of determining C being the closest route, the IGP might
> determine route A to be closer. Router D then sends packets
> destined for 10.0.0.0 to A. Router A receives these packets, and
> finds in its table the closest route is in the direction of C.
> Router A sends these packets to router D. The packets are caught
> in a routing loop.
>
> That's why I'm thinking redistribution must be restricted in this
> scenario. EBGP routes must be redistributed into the IGP at the
> same point where they are received from the EGP peer. But then my
> undestanding of BGP isn't that solid, and I haven't found
> anything written that concerns this scenario.
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Thu Dec 02 2004 - 06:57:49 GMT-3