North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

RE: IS-IS protocol implementation problem

  • From: rdobbins
  • Date: Mon Oct 30 17:53:11 2000

You're right - I'm looking for fault here in my own equipment, whereas a
screwed-up IGP on their end could well have caused all the problems I was

Occam's Razor, of course!

Your point about the iBGP confederations is well-taken, by the way.

Roland Dobbins <[email protected]> // 818.535.5024 voice 

-----Original Message-----
From: [email protected] [mailto:[email protected]]
Sent: Monday, October 30, 2000 11:11 AM
To: [email protected]
Subject: RE: IS-IS protocol implementation problem

[not to list]

| No, I'm a single-AS hosting provider, no confederation.  

UUNET has confederations; I was doing a public thinking exercise
(trying to coopt smart people who still read the NANOG list, too)
about the extent to which the insulation of iBGP in a confederated
AS from the disappearance of iBGP peers for an otherwise full-mesh
iBGP layout interacts with the exposure of multi-AS-one-IGP confederations
to failures which cause the iBGP next-hops to disappear.

If it seems a bit esoteric, don't worry, it is...

| The more I think
| about it, the more I'm convinced that CEF simply stopped working; all my
| interfaces were active, and there were no apparent problems with my IGP,
| which is OSPF.

Right, UUNET was having the problems; you were just a victim of
their internal routing being so broken that they couldn't make packets
move to you reliably, even though their routers knew how to get to
your network; likewise their routers kept telling you they knew how
to get to all sorts of networks which in fact they couldn't reach.

This problem appears consistent with an IGP problem inside UUNET,
which is known to use/have-used confederated ASes.

| I think that major BGP wigginess caused the CEF problem; thanks very much
| for you insight, I definitely need to think about it some more.

What makes you think it was a CEF problem?