Lightning Talks «Flight Recorder: Go's Black Box for Production»
Доклад принят в программу конференции
Целевая аудитория
Тезисы
Debugging production latency is an information problem. By the time you know something went wrong, the execution that caused it is gone. Traditional profiling asks you to predict when to look. FlightRecorder asks a better question: what if you were always looking?
FlightRecorder writes trace events into a ring buffer at 1-2% CPU overhead, goroutine state changes, blocking, mutex contention, and scheduling delays. When a request breaches P99 or an error spike hits, you snapshot. This talk covers FlightRecorder's architecture, why pprof can't explain most latency, the MinAge/MaxBytes tradeoff, and automatic snapshot patterns from production deployments.
Alex is a Principal Engineer at Memed, where he builds developer platforms and internal tools that empower engineering teams across the organization. With 17+ years of experience, he's the author of System Programming Essentials with Go and Learning Zig, and writes about staff engineering and systems thinking on Substack and his personal blog.
Alex speaks regularly at international conferences and is passionate about data-oriented design, making complex systems understandable, and helping engineers grow into technical leadership roles.
Видео
Другие доклады секции
Инструменты и фреймворки
