[6965] in North American Network Operators' Group
Re: withdrawal propagation (was E.E. Times?)
daemon@ATHENA.MIT.EDU (John W. Stewart III)
Tue Jan 14 14:36:46 1997
To: jon@branch.net (Jon Zeeff)
Cc: young@mci.net (Jeff Young), wsimpson@greendragon.com, nanog@merit.edu
In-Reply-To: Your message of "Tue, 14 Jan 1997 12:46:49 EST."
<m0vkCwj-000JChC@aero.branch.com>
Date: Tue, 14 Jan 1997 14:34:19 EST
From: "John W. Stewart III" <jstewart@metro.isi.edu>
excessive rates of bona fide routing updates *can* be a
problem. it's called route flap. and we've got route
flap dampening to reduce the scope of such events
what we've been talking about very recently on this list
is the high rate of withdrawls that have been seen.
specifically, e.g., withdrawls from RouterA to RouterB
for networks that RouterA never announced to RouterB.
this is not a route flap .. it is just a superfluous
withdrawl and causes no operational problems. however,
some folks were tracking the number of withdrawls and
didn't like the large number, so the vendor was informed
and the code was changed. it's a good and appropriate
thing that the behavior was changed, but that doesn't
mean that it was a bug and doesn't mean that it was
causing any problems
/jws
>
> > 0) Is this a bug, does it cause any problem whatsoever?
>
> If I'm not mistaken, lots of routers have had performance problems
> caused by excessive rates of routing updates.
>
> Or did I misread various previous messages to this list?
>
> > > I've looked at the Cisco page, and a search on "BGP, withdrawals" does
> > > not find any mention of the bug fix release. So, I have some pointed
>
