go中提供了pprof包来做代码的性能监控，在两个地方有包：

• net/http/pprof
• runtime/pprof

使用 net/http/pprof 做WEB服务器的性能监控

import (

_ “net/http/pprof”

)

然后就可以在浏览器中使用http://localhost:port/debug/pprof/ 直接看到当前web服务的状态，包括CPU占用情况和内存使用情况等。

在 main 方法中增加

func main() {
    go func() {
        http.ListenAndServe("localhost:8082", nil)
    }()
}

可以在终端执行

go tool  pprof –text http://localhost:port/debug/pprof/heap  查看分析数据

下面是我的测试样例:

Fetching profile from http://localhost:port/debug/pprof/heap
Saved profile in /root/pprof/pprof.localhost:port.inuse_objects.inuse_space.274.pb.gz
20512.19kB of 20512.19kB total (  100%)
Dropped 68 nodes (cum <= 102.56kB)
      flat  flat%   sum%        cum   cum%
   20000kB 97.50% 97.50%    20000kB 97.50%  github.com/gorilla/websocket.newConn
  512.19kB  2.50%   100%   512.19kB  2.50%  runtime.malg
         0     0%   100%    20000kB 97.50%  github.com/gorilla/websocket.(*Upgrader).Upgrade
         0     0%   100%    20000kB 97.50%  github.com/gorilla/websocket.Upgrade
         0     0%   100%    20000kB 97.50%  main.wsHandler
         0     0%   100%    20000kB 97.50%  net/http.(*ServeMux).ServeHTTP
         0     0%   100%    20000kB 97.50%  net/http.(*conn).serve
         0     0%   100%    20000kB 97.50%  net/http.HandlerFunc.ServeHTTP
         0     0%   100%    20000kB 97.50%  net/http.serverHandler.ServeHTTP
         0     0%   100%    20000kB 97.50%  runtime.goexit
         0     0%   100%   512.19kB  2.50%  runtime.mcommoninit
         0     0%   100%   512.19kB  2.50%  runtime.mpreinit
         0     0%   100%   512.19kB  2.50%  runtime.rt0_go
         0     0%   100%   512.19kB  2.50%  runtime.schedinit