Troubleshooting Periodic 100% CPU on a Bun Service
A few days ago we hit a fairly interesting production issue. A Bun service was periodically pegging one CPU core across three pods, for exactly 10 minutes each time, then dropping back on its own. The service itself was fine: requests worked, health checks never failed. The CPU graph just looked awful.
It turned out to be two problems that are each pretty mild on their own, stacked on a specific kernel version and a specific Bun patch release. Writing it down.