Earlier today I had an issue where a circuit to one of my two BGP connected upstreams went away for an hour or so. During this period, I expected BGP to act as expected and migrate the traffic to the second circuit with a second provider. This did not occur. Initially I figured this had to do with route flap amplification or similar causing route dampening. However, when the circuit came back up connectivity was almost immediately restored to the entire internet, which doesn't seem consistant with a route flap dampening, unless the timing was just coincidental. This leads me to believe that the routes may have not been withdrawn for the path through the second provider even though the circuit was down for ~90 minutes. How this would have occured I have no idea. At this point, I'm trying to reconstruct the state of the global routing table in relation to my prefixes during this period. I seem to recall at least historically that there was at least one or two places which were capturing route announcement/withdrawl data on the internet. However, google fails me. Is this data currently being captured anywhere, and if so, is this data publically available (or at least are the data "owners" willing to provide an extract for my prefixes)? Any pointers would be helpful. Thanks.
Talk to the renesys guys. Regards, Neil. ----- Any pointers would be helpful. Thanks.
On Thu, May 24, 2007 at 01:42:32AM -0600, Forrest W. Christian wrote:
However, google fails me. Is this data currently being captured anywhere, and if so, is this data publically available (or at least are the data "owners" willing to provide an extract for my prefixes)?
http://www.ris.ripe.net/perl-risapp/risearch.html would probably give you what you're looking for. cheers, -- Erik Romijn RIPE NCC jr. software engineer http://www.ripe.net/ Information Services dept.
At this point, I'm trying to reconstruct the state of the global routing table in relation to my prefixes during this period. I seem to recall at least historically that there was at least one or two places which were capturing route announcement/withdrawl data on the internet.
route-views project ripe/ncc ris project
On 24-May-2007, at 03:42, Forrest W. Christian wrote:
Earlier today I had an issue where a circuit to one of my two BGP connected upstreams went away for an hour or so. During this period, I expected BGP to act as expected and migrate the traffic to the second circuit with a second provider. This did not occur.
When this has happened to me before, I have been suspicious about whether the upstream to whom the circuit broke was routing my nets down the dead circuit with static routes that for whatever reason (layer-2 obfuscation, etc) didn't go away when the link went down. This hasn't always been the answer, but sometimes it has; in several cases poking about within RIS (or interactively through a route-views router while the circuit was down) revealed that upstream in question was originating routes on my behalf while the circuit was down. Joe
participants (6)
-
Erik Romijn
-
Forrest W. Christian
-
Joe Abley
-
Neil J. McRae
-
Peter Walker
-
Randy Bush