Installer async-profiler avec Homebrew, MacPorts, Nix

Liens du projet

• Page d'accueil
• Docs amont
• Page du gestionnaire de paquets

Commandes et alias

• asprof
• jfrconv

Historique du projet

The project started in 2016 with an initial proof of concept for profiling Java without the safepoint bias that affected many traditional JVM profilers. The README describes the core design as combining HotSpot-specific stack collection with native facilities such as perf events, so it can show Java frames alongside native, kernel, GC, and JIT compiler activity.

Over time async-profiler grew from a CPU sampler into a broader observability utility. Official documentation now covers CPU, allocation, native memory, lock, wall-clock, hardware/software counter, container, Java Flight Recorder, FlameGraph, OTLP, pprof, and converter workflows.

Historique d'adoption

The project became a default recommendation in the JVM performance community because it is scriptable, produces portable output, and does not require a heavyweight GUI profiler. The package is distributed through Homebrew, MacPorts, and Nix in the supplied package metadata, while upstream publishes Linux and macOS binaries plus nightly builds from GitHub Actions.

Recent 4.x releases show continued production-oriented maintenance rather than archival status: v4.3 added native lock profiling, latency filtering, Prometheus-format metrics, and Java-agent/JMX control, while v4.4 added differential flame graphs and call-trace memory limits.

Modes d'utilisation

The normal package-manager workflow is to install the CLI and run `asprof` against a live JVM PID, for example saving a 30-second profile to an HTML flame graph. More advanced workflows attach it as a Java agent, run continuous loops, emit JFR/OTLP/pprof-compatible data, or convert recordings with the bundled converter tooling.

Pourquoi les passionnés de paquets s'y intéressent

async-profiler is one of those packages whose formula matters more than its size suggests: a native tool with JVM-specific privileges, OS-specific packaging, architecture-specific binaries, and output formats that plug into a large performance-debugging ecosystem. It is also a good example of a CLI package that package managers make discoverable to users who otherwise would only encounter it through conference talks or performance postmortems.

Chronologie

• 2016: Initial Git commit and proof-of-concept async profiler work.
• v1.x: Early public release series established the standalone profiler package.
• 2026: v4.3 added native lock profiling, latency filtering, Prometheus metrics, and Java-agent/JMX control.
• 2026: v4.4 added differential flame graphs and further FlameGraph/JFR/OTLP improvements.

Related projects

• The project sits near HotSpot, Linux perf events, Java Flight Recorder, Brendan Gregg-style FlameGraphs, OpenTelemetry, and pprof-style profiling workflows.

Sources

• Official git metadata from https://github.com/async-profiler/async-profiler.git
• https://github.com/async-profiler/async-profiler#readme
• https://github.com/async-profiler/async-profiler/releases.atom
• https://github.com/async-profiler/async-profiler/tree/master/docs