To: bgpd@merit.edu, routing-wg@ripe.net Subject: Re: More route flaps & inconsistent origins Date: Thu, 13 Oct 1994 22:58:15 +0100 From: Havard Eidnes <Havard.Eidnes@runit.sintef.no>
Hi,
again, here is a new route flap report. The villains from the top of the previous report have apparently fixed their problem, but a few ASes again manage to get the dubious position at the top of the list with more than 20.000 route flaps in the 24 hours from Oct 12 03:05 GMT to Oct 13 03:05 GMT.
Again I would like to draw to your attention the fact that some destinations are apparently being originated by more than one AS. These show up in the report with more then one origin. I was under the impression that thsi is an illegal or highly undesireable configuration; I would still like to hear some comments on this issue.
This is not illegal by the BGP RFC, but it is basicly frowned upon by operators and the BGPd community. This usually occurs when people are back-leaking redistributing IGP's into different BGP ASs.
If you were homed to the NSFNET in two different places, you needed unique AS's at each peering point. So you had to dump your nets into two different AS's. I realize this recently changed, but why deal with it now, when it will soon go away.
mf
It has been a while since ANS/NSF needed multiple ASes on different peering points, however, I don't think ANS/NSF peering, or lack thereof, is the real issue, as you also alluded to. The important point is that the there are many incorrect home ASes tagged onto route announcements.
From an operational point of view, imho, it is helpful to be in a state where the home AS/network-prefix information is correct and consistent, both in the route annoucements and in the registration. There are cases where playing games w/route preferences based on home AS is required, and if they are not reliable, what good are they? I'd much rather deal w/AS path decisions than network prefixes. Make sense?
By the way, does anyone have a feel for how "good" the home AS registration information is/will be? -- Becca
By the way, an interesting statistic here is not so much the originating AS (yes, that is interesting, but...) but rather determining transit ASs that are causing the flaps.
For instance, say AS 109 is transited by AS 200, and AS 200 flaps a lot. AS 109 will show up as flapping and the flaps aren't debited against 200, so you don't discover the real problem.
Paul
* AS's used in this are only for example purposes, no slam against barrnet intended. :-)