Import net/http/pprof for web apps to enable profiling endpoints like /debug/pprof/profile and /debug/pprof/heap. 2. Use go tool pprof to collect and analyze profiles with commands like top, list, and web for CPU and memory data. 3. For non-HTTP apps, use runtime/pprof to write CPU and heap profiles to files and analyze them offline. 4. Interpret key metrics such as flat/cum CPU values, inuse_space in heap, and block/mutex contention to identify bottlenecks. 5. Visualize profiles using go tool pprof -web after installing graphviz, and generate SVGs for sharing. Profiling Go applications is built-in, efficient, and essential for performance optimization when done under realistic conditions with secure endpoint access.

Profiling Go applications is essential for identifying performance bottlenecks like CPU hogs, memory leaks, or excessive garbage collection. Go provides built-in tools through the pprof package that make profiling straightforward. Here's how to do it effectively.

1. Use net/http/pprof for Web Applications

If your Go application is a web server, the easiest way to enable profiling is by importing net/http/pprof. This automatically adds several debugging endpoints to your HTTP server.

import _ "net/http/pprof"
import "net/http"

func main() {
    go func() {
        log.Println(http.ListenAndServe("localhost:6060", nil))
    }()
    // your app logic
}

Once enabled, visit http://localhost:6060/debug/pprof/ in your browser to see available profiles:

• /debug/pprof/profile – CPU profile (30-second default)
• /debug/pprof/heap – Heap memory allocation
• /debug/pprof/goroutine – Goroutine stack traces
• /debug/pprof/block – Goroutine blocking profile
• /debug/pprof/mutex – Mutex contention

You can also fetch these profiles via curl or go tool pprof.

2. Collect and Analyze Profiles with go tool pprof

After enabling pprof, collect a CPU profile:

go tool pprof http://localhost:6060/debug/pprof/profile

This downloads a 30-second CPU profile and opens an interactive session.

For memory:

go tool pprof http://localhost:6060/debug/pprof/heap

Common pprof commands in the interactive shell:

• top – Show top functions by resource usage
• list <function> – Disassemble a specific function
• web – Generate a visual call graph (requires graphviz)
• trace – Output trace of matching samples
• text – Print flat text report

You can also skip the shell and generate output directly:

go tool pprof -top http://localhost:6060/debug/pprof/heap

3. Profile Non-HTTP Applications

For CLI or background apps without HTTP servers, use the runtime/pprof package directly.

CPU Profiling

f, _ := os.Create("cpu.prof")
pprof.StartCPUProfile(f)
defer pprof.StopCPUProfile()

// run the code you want to profile

Memory Profiling

// capture heap profile
f, _ := os.Create("heap.prof")
defer f.Close()
// do some work...
pprof.WriteHeapProfile(f)

Then analyze with:

go tool pprof cpu.prof
go tool pprof heap.prof

4. Interpret Key Metrics

• CPU Profile: Look for functions with high "flat" or "cum" values. These are consuming the most CPU time.
• Heap Profile: Check inuse_space for currently allocated memory. High values may indicate memory leaks.
• Goroutine Profile: Useful for spotting goroutine leaks or deadlocks.
• Block/Mutex Profiles: Reveal synchronization bottlenecks.

Tip: Use --nodefraction=0.05 in pprof to hide small nodes and focus on significant contributors.

Bonus: Use pprof with Visualization

Install graphviz (brew install graphviz or sudo apt-get install graphviz), then:

go tool pprof -web cpu.prof

This generates and opens a visual flame graph-style call tree.

You can also generate SVGs:

go tool pprof -web -output=profile.svg cpu.prof

Profiling Go apps doesn’t require external tools—just enable pprof, collect data under realistic load, and analyze. The key is to profile in an environment that mirrors production as closely as possible.

Basically, it’s built-in, lightweight, and powerful—just don’t forget to disable debug endpoints in production unless you lock them down.

The above is the detailed content of How to profile Go applications for performance?. For more information, please follow other related articles on the PHP Chinese website!