From: Bob Sinclair (bsin@xxxxxxxxx)
Date: Sun Mar 17 2002 - 13:12:47 GMT-3
Sandro,
I have to say that you make a lot of sense in your comments below, I have chang
ed my mind about the necessity for filtering when doing mutual redistribution.
If you are redistributing at just one point, and split-horizon is working, ther
e are many designs that do not seem to require filters. But there are many tha
t do, some of which are quite sneaky. One example is here, regarding OSPF forw
arding address getting screwed up:
http://www.cisco.com/warp/public/104/10.html
In that example, the need for filters on one box does not show up until you do
a summary on another box, then things start breaking, revealing the underlying
problem.
My thought had been that in a static, lab situation, why waste time on filters
(which are just another place for errors to pop up) if that particular design w
orks without them.
My present intention is to do the filters whether or not the design appears to
work without them, because 1) it is best practice and 2) the problems that coul
d pop up without them are insidious and difficult to troubleshoot.
Thanks for the input,
-Bob
----- Original Message -----
From: "Sandro Ciffali" <sandyccie@yahoo.com>
To: "Narvaez, Pablo" <Pablo.Narvaez@getronics.com>; "Bob Sinclair" <bsin@erols.
com>; "Landon Fitts" <l.fitts@mindspring.com>
Cc: <ccielab@groupstudy.com>
Sent: Saturday, March 16, 2002 7:28 PM
Subject: RE: Mutual Redistribution - OSPF / IGRP
> Hackito,
> I respectfuly disagree with you. Let us say you have a
> router running igrp and ospf both and is using subnets
> of a major class B network. Now if you redistribute
> ospf and igrp mutualy on this router, Remember igrp is
> a class full routing protocol and is running on all
> the interfaces of this router, (Even though in the
> passive mode). now when you redistribute between ospf
> and igrp without using distribute list you are taking
> advantage of the fact that ospf has a less admin
> distance as compared to igrp and hence everything is
> fine, Imagine ospf having larger admin distance than
> igrp, Now you will see all the ospf routes as type
> external instead of IA or internal.
> I think doing redistribution between any two protocols
> without reout-map or some kind of filtering has to be
> done with filtering no matter you think there is a
> route feedback problem or not. Ofcourse these are my
> thoughts, you have full rights to disagree with me.
>
> Sandro
>
> --- "Narvaez, Pablo" <Pablo.Narvaez@getronics.com>
> wrote:
> > I agree with Bob when saying "there is no need to
> > configure distribute-lists, route-maps or whatever
> > you want to use to avoid route feedback" .... ISn't
> > split-horizon supposed to take care of this issue?
> > ... The only case I can think of is when you have FR
> > in between R2 and R3; if you have a subint
> > point-to-point on R2 and you use the physical
> > interface on R3 you will have to worry about route
> > feedback on R2 since the physical interface on R3
> > will have split-horizon disabled ... now you have to
> > choose whether you want to use any kind of filter on
> > R2 or just enable splithorizon on R3 ...
> >
> > Please correct me if wrong ... besides, I would like
> > to know when I have to worry about route feedback in
> > this kind of scenario, with only one way in and out
> > for the routes ...
> >
> > Cheers,
> >
> > -hockito-
> >
> >
> >
> > -----Original Message-----
> > From: Bob Sinclair [mailto:bsin@erols.com]
> > Sent: Jueves, 14 de Marzo de 2002 09:46 p.m.
> > To: Landon Fitts
> > Cc: ccielab@groupstudy.com
> > Subject: Re: Mutual Redistribution - OSPF / IGRP
> >
> >
> > Landon,
> >
> > I am sure you will get a lot of feedback on this,
> > but let me give you my two cents worth.
> >
> > The straightforward, text book answer to the problem
> > is to create summary routes to null 0 and then
> > redistribute these statics into igrp. But I assume
> > that option is ruled out.
> >
> > There has been some discussion of creating a second
> > ospf process, redistributing into that to create an
> > asbr on R2, then do summary-addresses that can be
> > redistributed into igrp. Problem I see here, given
> > the ground rules, is that summary address
> > automatically creates the null0 routes you are
> > trying to avoid. Same thing happens if you create
> > an additional ospf area on R2 and do area-range:
> > recent IOS automatically creates summaries to null0.
> >
> > Contrived approach of secondaries does not scale,
> > but is only approach that could get the routes there
> > with no statics. At least that I see.
> >
> > My question is: given this scenario, do we have to
> > be concerned about route feedback? Why doesn't
> > split horizon take care of this? Or are we to
> > assume the network might somehow change?
> >
> > -Bob
> >
> > ----- Original Message -----
> > From: "Landon Fitts" <l.fitts@mindspring.com>
> > To: <ccielab@groupstudy.com>
> > Sent: Thursday, March 14, 2002 10:05 PM
> > Subject: Mutual Redistribution - OSPF / IGRP
> >
> >
> > > I was working on a mutual redistribution scenario
> > today between OSPF and
> > > IGRP, and a thought came to mind.
> > >
> > > Here is my scecnario:
> > >
> > > R1 ------------------ R2-------------------- R3
> > > OSPF OSPF | IGRP
> > IGRP
> > >
> > > My OSPF domain has networks that are /30, /29,
> > /28,and /24.
> > > My IGRP domain has only /24 networks.
> > > When I do mutual redistribution at R2 (and use
> > distribute-list to prevent
> > > route feedback) R1 sees all my IGRP networks, but
> > IGRP only sees
> > > my /24 networks from OSPF. So, to fix this I used
> > the "IP Default-Network"
> > > command at R3 pointing to a reachable loopback
> > interface
> > > in R2 that is part of my ospf domain. Now,
> > everybody is happy and all
> > > networks are reachable from each router.
> > >
> > > Now, what I want to know is is "ip
> > default-network" considered a static
> > > route? If so, what are the other alternatives to
> > gain reachability from R3
> > > to my other networks that are part of OSPF
> > domain. One other possible
> > > solution could be to use secondary address between
> > R2 and R3, but this isn't
> > > scalable because I have several subnets in OSPF.
> > >
> > > Regards,
> > >
> > > Landon Fitts
> > > NNCSE, NNCDE, CCNP, CCDP, CCIE Lab Candidate
> > > l.fitts@mindspring.com
> > >
> >
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:57:11 GMT-3