Skip to content

Latest commit

 

History

History
96 lines (66 loc) · 2.88 KB

README.md

File metadata and controls

96 lines (66 loc) · 2.88 KB
Raw

Profiler Utilities

Important Notes

  • This package is designed for troubleshooting performance related issues such as
    • Memory Leak
    • High CPU Usage
  • This package MUST NOT be used in production.

How To Use

This package leverage GO runtime/pprof and net/http/pprof packages.

For more efficient troubleshooting, this package should be used together with GO benchmarking techniques and suitable pprof visualization tools.

Serve pprof Data via HTTP

To enable pprof over HTTP, add following code before application start, usually in init() of application's main package:

package main
import "github.com/cisco-open/go-lanai/pkg/profiler"
func init() {
	profiler.Use()
}

The profiler package will enable runtime/pprof and install following endpoints under application's server.context-path:

  • GET /${server.contex-path}/debug/pprof/[pprof_profile]: Dump compressed raw data as gz.

    Common [pprof_profile] values are

    • heap
    • block
    • goroutine
    • threadcreate
    • mutex

  • GET /${server.contex-path}/debug/pprof/cmdline: Returns running program's command line, with arguments separated by NUL bytes.

  • GET /${server.contex-path}/debug/pprof/profile: Returns the pprof-formatted cpu profile.

  • GET /${server.contex-path}/debug/pprof/symbol: Looks up the program counters listed in the request, responding with a table mapping program counters to function names

  • GET /${server.contex-path}/debug/pprof/trace: Returns the execution trace in binary form

More details about GO pprof can be found at:

Visualize pprof Data

There are many tools available to visualize data produced by pprof. Here is an exmaple:

Tool: Google pprof GO implementation

Prerequisites:

  1. Go development kit (recommended 1.16+)

  2. Graphviz. To install on MacOS:

    brew install graphviz

Install google/pprof CLI:

go install github.com/google/pprof@latest

Example Usage (Investigating AuthService's memory allocation):

pprof -http=localhost:6061 http://localhost:8900/auth/debug/pprof/heap

Command above would start a web service at localhost:6061 and render various views of current heap snapshot.

Use pprof --help for detailed usage.

Real-Time Performance Monitoring

This folder also includes a performance monitoring utility that serve an HTML page of real-time metrics charts. To enable it:

package main
import "github.com/cisco-open/go-lanai/pkg/profiler/monitor"
func init() {
	monitor.Use()
}

The charts can be accessed via HTTP http://server:port/<context-path>/debug/charts/

Note: This is an experimental feature and may consume fare amount of resources.