------- Forwarded Message
Date: Mon, 28 Apr 1997 11:20:17 -0400 (EDT)
From: Andrew Partan <asp@partan.com>
Sender: owner-nanog@merit.edu
To: SEAN@SDG.DRA.COM (Sean Donelan)
cc: nanog@merit.edu
Subject: Re: MAI.net filters.> Any luck tracking down the reason why some routers inside Sprint kept
> announcing the routes long after the source was removed during either
> the incident last Friday or the other incident on Feb 3-4, 1997?Part 2 of the problem (where all of the routes didn't go away even
after they were removed by the source) wasn't only w/in Sprint; it
was elsewhere as well. This has been tracked down to a cisco bug,
which has been found & they are working on a fix.
--asp@partan.com (Andrew Partan)
More specifically this bug is relevant to EBGP peergroup members.
When there is a lot of churn, some of the EBGP peergroup members might
not receive the withdrawn message. This will happen only if the EBGP
neighbors have tons of update queued (ie unless the EBGP neighbors are
receiving thousands of prefixes, this bug will not be seen.). Thus it
is most likely to affect people using EBGP peergroups within the
confederation.
If you are not using EBGP peergroup, it will not have any problem. I
already have the fix and should be available soon. The bugid is
CSCdj08110.
--ravi