RE: [nsp] OSPF not distributing 1 interface

From: Chris Whyte (cwhyte@microsoft.com)
Date: Mon May 28 2001 - 22:59:58 EDT


>
> By what other method, short of a static routes, which to a dynamic
> protocol are just as bad, do you insert routes for all your connected
> interfaces? IMHO it's a very useful comand when you want all your
> interfaces to be advertised but do not necessarily run OSPF on all of
> them. Or, for that matter, interfaces that run other IGPs.
>

This has a potentially long answer since it depends on the problem
you're trying to solve. I would ask, why don't you want to run ospf on
those interfaces? Or, is there something wrong with running them
passive? Or, if those interfaces are running other igps then why not
just redistribute that igp into ospf?

What I highly suspect is that any perceived requirement to use this
command is probably just covering up a more fundamental architectural
issue. Therefore, I would suggest that the architectural issue be dealt
with first. I've never known anyone say that they need to use this
command when they finish creating configs to represent their original
architectural planning. And unfortunately when people do use it for the
"one-offs" in their network it has a tendency to set a precedent to use
it all over the place.
 
> I've never heard of this being an issue, I'd be interested to hear why
> it is such an "evil" command?

The evil is more in regard to the precedent I mentioned since using it
all over the place can increase the size of the LSDB considerably. But,
more specifically, you've now made the router an ASBR which it didn't
ncessarily need to be. This, of course, causes your abrs to generate
summary-asbrs for those routers. Any interface that had a network
statement covering it now essentially has a duplicate entry represented
by an as-external. And this somewhat defeats the purpose of hierarchy
and summarization.

There's probably some more I'm missing but that's just off the top of my
head without really thinking about it that much.

Thanks,

Chris

>
> -David Curran
>
> On 24 May 2001 06:04:53 -0400, Chris Whyte wrote:
> > I just shudder at the fact that people are even thinking about using
> > 'redist connected' as an option to get routes associated
> with connected
> > interfaces into OSPF. Bug or not, I've never been in a
> situation where I
> > had to result to using this approach. In large networks
> (since this is
> > the nsp mailing list) it's an evil command and really can't
> think of a
> > sane config that requires its use where there wasn't a better way of
> > getting those interfaces into your igp.
> >
> > I'd be quite happy if csco got rid of it...
> >
> > Thanks,
> >
> > Chris
> >
> > >
> > > > Having the interface IP match the network statement qualifies an
> > > > interface to run ospf, this is separate issue from passing that
> > > > specific connected route or other routes from the routing table
> > > > into ospf.
> > >
> > > It sounds like you are disagreeing with what Ken said, but
> > > what Ken said
> > > is correct. When a network statement matches an interface's
> > > subnet, that
> > > subnet will get dumped into the routing table as an intra
> > > area route. If
> > > you 'redistribute connected', you shove it in as an
> external route.
> > >
> > >
> >
>
>



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:39 EDT