Go 1.25 Ships Flight Recorder for Runtime Trace Diagnostics

Go 1.25 introduces the flight recorder, a new capability in the runtime/trace package built on the more powerful execution tracer first previewed in 2024. It lets programs buffer the last few seconds of execution trace data in memory and dump a snapshot on demand, right when a problem is detected.

Traditional Start/Stop tracing works fine for tests or short-lived tools, but it's impractical for web services that run for days or weeks. Fleet-wide random sampling is an alternative, but it demands heavy infrastructure to store and triage mostly uninteresting trace data. The flight recorder solves this by maintaining a configurable ring buffer (via MinAge and MaxBytes settings) and exposing a WriteTo method to capture exactly the problematic window once an issue surfaces.

The post walks through a concrete example: an HTTP server running a 'guess the number' game that occasionally responds in over 100 milliseconds instead of microseconds. By starting the flight recorder and triggering a snapshot on slow responses, engineers can capture the exact execution trace leading up to the latency spike, without instrumenting or tracing the entire service continuously. This gives Go developers a lightweight, production-safe way to root-cause intermittent performance issues.