Hi Rick,
Yes, what you have identified is indeed a problem...
When picking the flap damping parameters care is required, and the RIPE
value for /24 prefixes does not produce any damping at all. (And IOS
doesn't warn you about that problem either, unfortunately.)
set damp 30 750 3000 60 basically says half life is 30 minutes,
max-suppress is 60 minutes, reuse is penalty of 750. Worst case, penalty
will have decayed to 750 after 60 minutes for reuse - 30 minutes prior to
that, the penalty will have been 1500, and 30 minutes prior to that the
penalty will have been 3000. Penalty needs to be higher than the 3000
suppress limit before the prefix will be damped.
Folks using Cisco routers should make the reuse limit 800, say, (or the
suppress limit 2500) for damping to actually work.
A revision to RIPE-210 will be issued in due course.
cheers,
philip
--PS. The other problem with RIPE-210 is that "random" root and tld servers are listed in the exclusion list. The exclusion list should contain all important infrastructure, and folks should recognise that important infrastructure such as the roots will change IP address or announced network from time to time. Check all values before implementing.
At 19:00 19/10/00 +0200, Christian Panigl, ACOnet/VIX/UniVie wrote: > Rick, > > thank you for your notification regarding RIPE-210 flaws. We have > already identified this and some other problems, have been discussing > those issues during the last RIPE meeting and are working on an update > of this document. Philip Smith has volunteered to specifically > cross-check all Cisco related stuff therein. > > Thanks again and kind regards > CP > >========================================= >Date: Thu, 19 Oct 2000 12:41:17 -0400 >From: Rick Payne <rickp@rossfell.co.uk> >To: cisco-nsp@puck.nether.net >Subject: [nsp] BGP Dampening bug? > > >We've been running with the RIPE specified values on BGP dampening for a >while now. However - I noticed an oddity today when debugging zebra. > >A route which was flapping was getting to its supress value, but never >going above it. This meant the route was withdrawn but almost immediately >re-instated with no hold down. It turns out that this is because we had the >supress value as a multiple of the half life. > >We had: > > route-map graded-flap-damping permit 20 > match ip address prefix-list min24 > set damp 30 750 3000 60 > >This caused oscillations inside our mesh - because the route was never >getting held down. > >Once we changed it to > > route-map graded-flap-damping permit 20 > match ip address prefix-list min24 > set damp 30 750 3001 60 > >Then the route got held down properly. > >Is this a bug in our understanding? or the code. I guess that as RIPE >specify this - there will be lots of people using it. > >Rick > >--------------------------------------------------------------------------------
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:19 EDT