[114009] in North American Network Operators' Group
Re: route flap dampening
daemon@ATHENA.MIT.EDU (David Storandt)
Mon Apr 27 15:13:27 2009
In-Reply-To: <49F5FB06.8090804@brightok.net>
Date: Mon, 27 Apr 2009 15:13:09 -0400
From: David Storandt <dstorandt@teljet.com>
To: NANOG list <nanog@nanog.org>
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
We're using Cisco 6509 MSFC2s for core engines with three 85% route
feeds and Cisco-default route flap suppression. Yeah, the default flap
suppression parameters are aggressive but we want to be sure we don't
hog precious CPU cycles from a nasty route flap and provide more
consistent routes to our downstreams. We can't take a full route table
(232k) due to TCAM limitations, so we have default routes anyway.
There's only a subtle impact to our customers with maybe a less
preferable, stable path versus a better, flapping one. A better
bargain in our book, but maybe not for others... CPU runs around
10-12% all day.
If we had more router CPU and no default routes, we'd probably have
dampening enabled but at very high thresholds under similar network
policies.
-D
On Mon, Apr 27, 2009 at 2:35 PM, Jack Bates <jbates@brightok.net> wrote:
> Patrick W. Gilmore wrote:
>>
>> On Apr 27, 2009, at 2:20 PM, Jack Bates wrote:
>>>
>>> We've been considering it after the last flap around the world; perhaps
>>> with extremely short penalty times.
>>
>>
>> <http://www.ripe.net/ripe/meetings/ripe-43/presentations/ripe43-routing-=
flap.pdf>
>>
>
> Yeah, read the presentation several times, thus the short penalty times, =
and
> probably high thresholds.
>
> The idea for me is to limit the harm of excessive flapping while not bein=
g
> paranoid. I've had a customer lose a lot of connectivity for 30 minutes
> after 3 or 4 flaps. I figure 5 minute ignore after about 10 flaps in a =
=A010
> minute period.
>
> Jack
>
>
