datadog/go-profiler-notes

This repository is a collection of research notes on profiling Go programs. Profiling is the practice of measuring where a program spends its time or memory so developers can find and fix performance problems. Go has several built-in profiling tools, but the author found that documentation on how they actually work is thin, so they wrote up their own findings here. The notes were started by an engineer who joined Datadog to work on a continuous profiling product for Go programs. The content lives in separate documents covering different profiler types: a guide to profiling, tracing, and observability in Go overall, a walkthrough of the pprof tool and its data format, a deep look at how Go stack traces work, notes on goroutine profiling (which shows what every active thread in a Go program is doing at a moment in time), and notes on block profiling (which tracks how long code waits on communication channels and synchronization locks). Several more sections on CPU, heap, mutex, and wallclock profiling were marked as coming soon at the time of writing. The main README also links to external articles and Go documentation that the author found useful, covering topics like custom profiling labels, mutex profiling, and using macOS developer tools alongside Go's built-in profilers. This is a documentation and notes repository, not a library or tool you install and run. It is aimed at Go developers who want a deeper understanding of what the profiling output actually means, not just how to collect it. The markdown files are licensed under the Creative Commons Attribution-ShareAlike 4.0 license.