Re: IBGP and IGP interaction
kannan@catarina.usc.edu Fri, 26 May 1995 18:32 UTC
Received: from ietf.nri.reston.va.us by IETF.CNRI.Reston.VA.US id aa07257;
26 May 95 14:32 EDT
Received: from CNRI.Reston.VA.US by IETF.CNRI.Reston.VA.US id aa07253;
26 May 95 14:32 EDT
Received: from interlock.ans.net by CNRI.Reston.VA.US id aa12685;
26 May 95 14:32 EDT
Received: by interlock.ans.net id AA41803
(InterLock SMTP Gateway 3.0 for iwg-out@ans.net);
Fri, 26 May 1995 14:04:16 -0400
Message-Id: <199505261804.AA41803@interlock.ans.net>
Received: by interlock.ans.net (Protected-side Proxy Mail Agent-2);
Fri, 26 May 1995 14:04:16 -0400
Received: by interlock.ans.net (Protected-side Proxy Mail Agent-1);
Fri, 26 May 1995 14:04:16 -0400
X-Authentication-Warning: catarina.usc.edu: Host catarina.usc.edu didn't use
HELO protocol
Sender: ietf-archive-request@IETF.CNRI.Reston.VA.US
From: kannan@catarina.usc.edu
To: Radha Gowda <rxg@proteon.com>
Cc: bgp@ans.net
Subject: Re: IBGP and IGP interaction
In-Reply-To: Your message of Fri, 26 May 1995 12:23:03
-0400.<199505261623.AA39278@interlock.ans.net>
Date: Fri, 26 May 1995 11:04:07 -0700
X-Orig-Sender: kannan@catarina.usc.edu
>>> From: rxg@proteon.com (Radha Gowda) >>> Date: Fri, 26 May 1995 12:23:03 EDT > I am a bit confused with IBGP and IGP interaction. I'd appreciate > any clarifications. > > AS 1 AS 2 > ----- ----- > | 1 | -----/---- | 2 | > ----- ----- > \ / > -----------\--------------/----------- > / \ AS 3 > \ / > ----- ----- > | 4 |--/--| 3 | (the connection between 3 and 4 could be > ----- ----- direct or many hops inbetween) > > IBGP and OSPF/RIP > > > IBGP is used only for exchanging routing information across an AS, and > the updates should not be used to install routes in the forwarding table. > This makes sense, if the AS is a transit AS and with automatic tag > generation, the OSPF external routes will not even be considered while > sending updates to external neighbors. > > Suppose there is already a BGP readvertisable route (learnt from EBGP) > in the forwarding table, and we learn an alternate path from IBGP > that we consider as well although not the best path. What is the correct > behaviour when the external path goes away? What do you want to do when the external route goes away? Do you want to re-advertise a new route to your external peers? Or do you want to install a route in your forwarding table only? The latter is a pre-requisite for the former. If you want to re-advertise a route learned via IBGP, then you must have an IGP learned route to the destination installed in your forwarding table, before you re-advertise the route to your external peers. If the route is imported into at most one ASBR, then this is not a problem. If you have more than ASBR importing the same route into your IGP, then you must be able to decide the route through that ASBR that is installed in your forwarding table via your IGP, and re-advertise that route. In the case of OSPF with automatic tag generation, this is outlined in section 3 of RFC1745. What the RFC does not say (and now that I think about it, wish it had done so), is that, in the event that you are not able to un-equivocally decide which ASBR is being used to reach a particular destination, either because you are using OSPF without automatic tag generation, or a brain dead IGP like RIP, then you must merge the PATH advertisements for all ASBRs that have imported the route into their IGP. See the example in the appendix section 10, for why this is necessary. > Or suppose the path learnt from IBGP is a better path. So should the > BGP route be deleted from the forwarding table? Or should it be deleted > only if an internal path exists to the route? As long as you don't re-advertise a route to other peers, then you could install anything you like :-). The sane thing might be to delete the external route and install the route learned via IBGP only if you have a matching IGP learned route as well. (In a lot of cases, that may be the only thing you can do...) Kannan
- IBGP and IGP interaction Radha Gowda
- Re: IBGP and IGP interaction kannan
- Re: IBGP and IGP interaction Radha Gowda