To: nwg@ncar.ucar.edu
cc: dsg@ncar.ucar.edu, sitongia@hao.ucar.edu
Subject: Routing and todays problems
Date: Mon, 16 Sep 1996 16:50:40 -0600
From: Paul Hyder <hyder@triceratops.scd.ucar.edu>

First:  This doesn't happen very often.  For that I'm thankful.

Now, since there seems to be some confusion, on to what actually
happens when we lose default route...

When we lose the link to the outside, default route (correctly)
goes away.  (As happened today.)

This means that statically default route configured hosts trying to
reach the outside will get a "network is unreachable" from the first
router.  This is also correct, we don't want to use up bandwidth
traversing the network to just to fail.

All of the local routers talk OSPF.  Anything talking OSPF
keeps the full set of active local routes.  So what happened
to cause problems with hosts and local connectivity?  Bluntly,
it was mostly routed!

For the record:  None of the routers lost internal routes.

Hosts that only talk RIP lost their default route.  Since they
don't have full routing this means that they lose some or all
internal connectivity.  The problem with routed is it's lack
of the ability to add static routes.

The solutions to this rare problem remain:

	1.  Get a secondary outside link via CSN.  (!In process!)
	2.  All hosts that must run dynamic routing should be
	    running gated with OSPF.  [Static routes actually DO
	    work in most cases.]

	And since we don't seem to have had much success with #2:

	A.  Convince folks to absolutely avoid routed.
	B.  If a host must run RIP based dynamic routing, use gated
	    and configure in a high metric default route.

As always life continues to be interesting.  Let me know if you
have questions.
	Paul Hyder