Go内存泄漏 Golang Memory Leak
环境
- go v1.14.4
什么是 Pprof
首先，不得不提的就是go分析利器pprof。pprof记录程序在运行过程中的cpu使用情况、内存使用情况、 goroutine运行情况、阻塞状况等，是定位bug和性能分析的利器。
如何开启pprof
go中有两个地方有：
- net/http/pprof
- runtime/pprof
差别只是runtime/pprof通过封装暴露http端口后就是net/http/pprof
开启pprof很简单，只需导入net/http/pprof包，并开启一个goroutine去监听http端口6060就可以了，具体代码如下：
package main
import (
"fmt"
"net/http"
_ "net/http/pprof"
"github.com/gorilla/mux"
"time"
"log"
)
func hello(w http.ResponseWriter, r *http.Request) {
w.Write([]byte("hello world"))
}
func main() {
go func() {
ip := ":6060"
if err := http.ListenAndServe(ip, nil); err != nil {
fmt.Printf("start pprof failed on %s\n", ip)
}
}()
r := mux.NewRouter()
r.HandleFunc("/hello", hello)
srv := &http.Server{
Handler: r,
Addr: ":8080",
WriteTimeout: 15 * time.Second,
ReadTimeout: 15 * time.Second,
}
log.Fatal(srv.ListenAndServe())
}
运行之后，可以在浏览器中打开ip:6060/debug/pprof页面，其中的信息大致如下：
- allocs：所有过去内存分配的抽样
- block：groutine阻塞的堆栈信息
- cmdline：当前程序的命令行调用
- groutine：所有goroutine的信息
- heap: 活动对象的内存分配抽样。在获取堆样本之前，可以指定gc GET参数来运行gc
- mutex: 争用互斥锁持有者的堆栈跟踪
- profile: CPU配置文件。您可以在seconds GET参数中指定持续时间。获取概要文件后，使用go tool pprof命令来调查概要文件。
- threadcreate: 导致创建新操作系统线程的堆栈跟踪
- trace: 当前程序的执行痕迹。您可以在seconds GET参数中指定持续时间。获取跟踪文件后，使用go tool trace命令来调查跟踪。
- full groutine stack dump: goroutine调用栈信息
命令行交互方式访问
# 可以获取指定的profile文件，此命令会发起http请求，并下载数据到本地，之后进入交互模式
go tool pprof url
# url可以是：http://127.0.0.1:6060/debug/pprof/heap之类
# 进入交互终端后，可以执行help查看命令说明
# 如：go tool pprof http://127.0.0.1:6060/debug/pprof/goroutine之后运行top命令，结果如下：
(pprof) top
Showing nodes accounting for 4, 100% of 4 total
Showing top 10 nodes out of 26
flat flat% sum% cum cum%
2 50.00% 50.00% 2 50.00% runtime.gopark
1 25.00% 75.00% 1 25.00% net/http.(*connReader).backgroundRead
1 25.00% 100% 1 25.00% runtime/pprof.writeRuntimeProfile
0 0% 100% 2 50.00% internal/poll.(*FD).Accept
0 0% 100% 2 50.00% internal/poll.(*FD).acceptOne
0 0% 100% 2 50.00% internal/poll.(*ioSrv).ExecIO
0 0% 100% 2 50.00% internal/poll.(*pollDesc).wait
0 0% 100% 2 50.00% internal/poll.runtime_pollWait
0 0% 100% 1 25.00% main.main
0 0% 100% 1 25.00% main.main.func1
# - flat: 给定函数上运行goroutine数量
# - flat%: 给定函数上goroutine运行数量总占比
# - sum%: 给定函数累计goroutine数量总占比
# - cum: 当前函数加上它之前的调用中goroutine运行总数量
# - cum%: 当前函数及之前调用中goroutine数量总占比
# 最后一列是函数名称
运行go tool pprof http://127.0.0.1:6060/debug/pprof/heap命令可以加上一些参数：
- -inuse_space: 分析应用程序的常驻内存占用情况，通常用来检测有没有不符合预期的内存 对象引用
- -alloc_objects: 分析应用程序的内存临时分配情况
# go tool pprof http://127.0.0.1:6060/debug/pprof/goroutine之后运行list main命令，结果如下：
(pprof) list main
Total: 4
ROUTINE ======================== main.main in F:\shared\gopath\src\test_pprof\main.go
0 1 (flat, cum) 25.00% of Total
. . 30: Addr: ":8080",
. . 31: WriteTimeout: 15 * time.Second,
. . 32: ReadTimeout: 15 * time.Second,
. . 33: }
. . 34:
. 1 35: log.Fatal(srv.ListenAndServe())
. . 36:}
ROUTINE ======================== main.main.func1 in F:\shared\gopath\src\test_pprof\main.go
0 1 (flat, cum) 25.00% of Total
. . 16:}
. . 17:
. . 18:func main() {
. . 19: go func() {
. . 20: ip := ":6060"
. 1 21: if err := http.ListenAndServe(ip, nil); err != nil {
. . 22: fmt.Printf("start pprof failed on %s\n", ip)
. . 23: }
. . 24: }()
. . 25:
. . 26: r := mux.NewRouter()
ROUTINE ======================== runtime.main in E:\Go\src\runtime\proc.go
0 1 (flat, cum) 25.00% of Total
. . 198: // A program compiled with -buildmode=c-archive or c-shared
. . 199: // has a main, but it is not executed.
. . 200: return
. . 201: }
. . 202: fn := main_main // make an indirect call, as the linker doesn't know the address of the main package when laying down the runtime
. 1 203: fn()
. . 204: if raceenabled {
. . 205: racefini()
. . 206: }
. . 207:
. . 208: // Make racy client program work: if panicking on
list main查看main函数的代码以及该函数每行代码的指标信息，当然list参数可以进行模糊匹配。
# traces命令可以打印所有调用栈信息
(pprof) traces
Type: goroutine
Time: Apr 9, 2021 at 2:46pm (CST)
-----------+-------------------------------------------------------
1 runtime.gopark
runtime.netpollblock
internal/poll.runtime_pollWait
internal/poll.(*pollDesc).wait
internal/poll.(*ioSrv).ExecIO
internal/poll.(*FD).acceptOne
internal/poll.(*FD).Accept
net.(*netFD).accept
net.(*TCPListener).accept
net.(*TCPListener).Accept
net/http.(*Server).Serve
net/http.(*Server).ListenAndServe
main.main
runtime.main
-----------+-------------------------------------------------------
1 runtime.gopark
runtime.netpollblock
internal/poll.runtime_pollWait
internal/poll.(*pollDesc).wait
internal/poll.(*ioSrv).ExecIO
internal/poll.(*FD).acceptOne
internal/poll.(*FD).Accept
net.(*netFD).accept
net.(*TCPListener).accept
net.(*TCPListener).Accept
net/http.(*Server).Serve
net/http.(*Server).ListenAndServe
net/http.ListenAndServe
main.main.func1
-----------+-------------------------------------------------------
1 net/http.(*connReader).backgroundRead
-----------+-------------------------------------------------------
1 runtime/pprof.writeRuntimeProfile
runtime/pprof.writeGoroutine
runtime/pprof.(*Profile).WriteTo
net/http/pprof.handler.ServeHTTP
net/http/pprof.Index
net/http.HandlerFunc.ServeHTTP
net/http.(*ServeMux).ServeHTTP
net/http.serverHandler.ServeHTTP
net/http.(*conn).serve
-----------+-------------------------------------------------------
pprof可视化界面
首先，运行go tool pprof http://127.0.0.1:6060/debug/pprof/profile获取profile文件
启动可视化界面方法：
# 方法一
go tool pprof -http=:8080 profile
# 方法二
go tool pprof profile
(pprof) web
内存泄漏
使用监控工具可以发现：随着时间的推进，内存的占用率在不断的提高，这是内存泄露的最明显现象。