Suboptimal routing for a subset of prefixes from STO

时间线

1. 已解决2025年2月6日 17:26 UTC

   Routing is back to normal from every probe location. Duration: 13 minutes. Latency on the affected path was up to 15 ms above normal; no connectivity was lost.

2. 已确定原因2025年2月6日 17:20 UTC

   A transit provider two hops away leaked a more specific route for one of our /24s. We have announced the same more specific ourselves so that it wins on the shortest path, and we have asked GlobalConnect to filter the leak upstream.

3. 调查中2025年2月6日 17:16 UTC

   A subset of our STO prefixes is being reached over an unusually long path from Taiwan. We see it in our external probes; users in other regions are not affected.