BGP session instability with Cogent in ZRH

时间线

1. 已解决2024年7月27日 19:45 UTC

   Stable since 19:45 UTC. Outbound traffic is back on its normal paths. Duration: 37 minutes. Impact was limited to brief latency spikes on some routes; no session to a machine was reset.

2. 观察中2024年7月27日 19:36 UTC

   The session has been stable for 9 minutes. Traffic remains on Init7 while we watch.

3. 已确定原因2024年7月27日 19:18 UTC

   The resets coincide with a route-refresh storm on the Cogent side. We have applied dampening on our end and shifted outbound traffic to Init7 while Cogent investigates.

4. 调查中2024年7月27日 19:09 UTC

   The BGP session with Cogent in ZRH has reset 7 times since 19:08 UTC. Each reset causes a few seconds of reconvergence for traffic that prefers that path. We are looking at our side of the session first.