golang pprof 使用

一、概述

go的pprof工具可以用來監測進程的運行數據，用於監控程序的性能，對內存使用和CPU使用的情況統信息進行分析。

官方提供了兩個包：runtime/pprof和net/http/pprof，前者用於普通代碼的性能分析，后者用於web服務器的性能分析。

官方文檔：

https://golang.org/pkg/runtime/pprof/

https://golang.org/pkg/net/http/pprof/#Index

https://github.com/google/pprof/blob/master/doc/pprof.md

二、runtime/pprof的使用

該包提供了一系列用於調試信息的方法，可以很方便的對堆棧進行調試。

通常用得多得是以下幾個：

• StartCPUProfile：開始監控cpu。
• StopCPUProfile：停止監控cpu，使用StartCPUProfile后一定要調用該函數停止監控。
• WriteHeapProfile：把堆中的內存分配信息寫入分析文件中。

示例代碼

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

package main import ( "flag" "runtime/pprof" "log" "runtime" "math/rand" "os" "time" ) var cpuprofile = flag.String("cpuprofile", "", "write cpu profile `file`") var memprofile = flag.String("memprofile", "", "write memory profile to `file`") const ( col = 10000 row = 10000 ) func main() { flag.Parse() if *cpuprofile != "" { f, err := os.Create(*cpuprofile) if err != nil { log.Fatal("could not create CPU profile: ", err) } if err := pprof.StartCPUProfile(f); err != nil { //監控cpu log.Fatal("could not start CPU profile: ", err) } defer pprof.StopCPUProfile() } // 主邏輯區，進行一些簡單的代碼運算 x := [row][col]int{} s := rand.New(rand.NewSource(time.Now().UnixNano())) for i := 0; i < row; i++{ for j := 0; j < col; j++ { x[i][j] = s.Intn(100000) } } for i := 0; i < row; i++{ tmp := 0 for j := 0; j < col; j++ { tmp += x[i][j] } } if *memprofile != "" { f, err := os.Create(*memprofile) if err != nil { log.Fatal("could not create memory profile: ", err) } runtime.GC() // GC，獲取最新的數據信息 if err := pprof.WriteHeapProfile(f); err != nil { // 寫入內存信息 log.Fatal("could not write memory profile: ", err) } f.Close() } }

編譯運行，會根據入參生成兩個對應的數據文件：

1 2 3 4 5 6 7 8

> go build > ./pprof -cpuprofile cpu.prof -memprofile mem.prof > ll total 2656 -rw-r--r--. 1 ma root 1183 Jan 9 21:39 cpu.prof # cpu運行數據 -rw-r--r--. 1 ma root 280 Jan 9 21:39 mem.prof # 內存數據 -rwxr-xr-x. 1 ma root 2707088 Jan 9 21:39 pprof # 編譯后的可執行文件 -rw-r--r--. 1 ma root 1296 Jan 9 21:38 pprof.go

使用go tool pprof命令即可分析。

三、go tool pprof

生成數據文件后使用go tool pprof file進入交互式界面進行數據分析，輸入help可以查看命令。

用法詳見：pprof.md

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

> go tool pprof cpu.prof Entering interactive mode (type "help" for commands) (pprof) help Commands: cmd [n] [--cum] [focus_regex]* [-ignore_regex]* Produce a text report with the top n entries. Include samples matching focus_regex, and exclude ignore_regex. Add --cum to sort using cumulative data. Available commands: callgrind Outputs a graph in callgrind format disasm Output annotated assembly for functions matching regexp or address dot Outputs a graph in DOT format eog Visualize graph through eog evince Visualize graph through evince gif Outputs a graph image in GIF format gv Visualize graph through gv list Output annotated source for functions matching regexp pdf Outputs a graph in PDF format peek Output callers/callees of functions matching regexp png Outputs a graph image in PNG format proto Outputs the profile in compressed protobuf format ps Outputs a graph in PS format raw Outputs a text representation of the raw profile svg Outputs a graph in SVG format tags Outputs all tags in the profile text Outputs top entries in text form top Outputs top entries in text form tree Outputs a text rendering of call graph web Visualize graph through web browser weblist Output annotated source in HTML for functions matching regexp or address peek func_regex Display callers and callees of functions matching func_regex. ...

1.top

命令格式：top [n]，查看排名前n個數據，默認為10。

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

(pprof) top 8490ms of 8510ms total (99.76%) Dropped 13 nodes (cum <= 42.55ms) Showing top 10 nodes out of 15 (cum >= 110ms) flat flat% sum% cum cum% 6780ms 79.67% 79.67% 8510ms 100% main.main 670ms 7.87% 87.54% 1250ms 14.69% math/rand.(*Rand).Int31n 350ms 4.11% 91.66% 1600ms 18.80% math/rand.(*Rand).Intn 260ms 3.06% 94.71% 580ms 6.82% math/rand.(*Rand).Int31 190ms 2.23% 96.94% 190ms 2.23% math/rand.(*rngSource).Int63 130ms 1.53% 98.47% 320ms 3.76% math/rand.(*Rand).Int63 110ms 1.29% 99.76% 110ms 1.29% runtime.memclrNoHeapPointers 0 0% 99.76% 8510ms 100% runtime.goexit 0 0% 99.76% 110ms 1.29% runtime.heapBits.initSpan 0 0% 99.76% 110ms 1.29% runtime.largeAlloc

2.tree

命令格式：tree [n]，以樹狀圖形式顯示，默認顯示10個。

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

(pprof) tree 5 8250ms of 8510ms total (96.94%) Dropped 13 nodes (cum <= 42.55ms) Showing top 5 nodes out of 15 (cum >= 190ms) ----------------------------------------------------------+------------- flat flat% sum% cum cum% calls calls% + context ----------------------------------------------------------+------------- 6780ms 79.67% 79.67% 8510ms 100% | main.main 1600ms 100% | math/rand.(*Rand).Intn ----------------------------------------------------------+------------- 1250ms 100% | math/rand.(*Rand).Intn 670ms 7.87% 87.54% 1250ms 14.69% | math/rand.(*Rand).Int31n 580ms 100% | math/rand.(*Rand).Int31 ----------------------------------------------------------+------------- 1600ms 100% | main.main 350ms 4.11% 91.66% 1600ms 18.80% | math/rand.(*Rand).Intn 1250ms 100% | math/rand.(*Rand).Int31n ----------------------------------------------------------+------------- 580ms 100% | math/rand.(*Rand).Int31n 260ms 3.06% 94.71% 580ms 6.82% | math/rand.(*Rand).Int31 190ms 100% | math/rand.(*rngSource).Int63 ----------------------------------------------------------+------------- 190ms 100% | math/rand.(*Rand).Int31 190ms 2.23% 96.94% 190ms 2.23% | math/rand.(*rngSource).Int63 ----------------------------------------------------------+-------------

3.web

以web形式查看，在web服務的時候經常被用到，需要安裝gv工具，官方網頁：http://www.graphviz.org/。

linux用戶使用yum install graphviz安裝即可，當然，純命令行界面是不能查看的。

windows用戶下載msi包安裝后需要把安裝目錄下的bin目錄添加到環境變量才行。

如果沒有安裝gv工具，使用會報錯：

Cannot find dot, have you installed Graphviz?

exec: “firefox”: executable file not found in $PATH

4.其他

其他的都是以不同形式展現出來，大同小異，以后有時間再測試。

四、web服務器監測

在web服務器中監測只需要在import部分加上監測包即可：

1 2 3

import( _ "net/http/pprof" )

當服務開啟后，在當前服務環境的http://ip:port/debug/pprof頁面可以看到當前的系統信息：

點擊查看具體的信息：

通常可以對服務器在一段時間內進行數據采樣，然后分析服務器的耗時和性能:

1	go tool pprof http://*:*/debug/pprof/profile

使用該命令后會對服務進行30s的采樣，這段時間內可以盡量多使用web服務，生成多一些統計數據。

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

> go tool pprof http://127.0.0.1:8080/debug/pprof/profile Fetching profile from http://127.0.0.1:8080/debug/pprof/profile Please wait... (30s) Saved profile in \pprof\pprof.127.0.0.1.samples.cpu.001.pb.gz Entering interactive mode (type "help" for commands) (pprof) top 3870ms of 4800ms total (80.62%) Dropped 37 nodes (cum <= 24ms) Showing top 10 nodes out of 66 (cum >= 110ms) flat flat% sum% cum cum% 1230ms 25.62% 25.62% 1300ms 27.08% runtime.mapaccess1_faststr 860ms 17.92% 43.54% 860ms 17.92% runtime.memclrNoHeapPointers 810ms 16.88% 60.42% 1010ms 21.04% runtime.scanobject 190ms 3.96% 64.38% 190ms 3.96% runtime.heapBitsForObject 160ms 3.33% 67.71% 190ms 3.96% strconv.ParseInt 140ms 2.92% 70.62% 1720ms 35.83% business_sets/haoxingdai_qiangdan/server/handler.makeOrder4Replace 140ms 2.92% 73.54% 1990ms 41.46% runtime.mallocgc 120ms 2.50% 76.04% 120ms 2.50% runtime.heapBitsSetType 110ms 2.29% 78.33% 1680ms 35.00% runtime.mapassign 110ms 2.29% 80.62% 110ms 2.29% runtime.memhash

使用web命令后會生成采樣時間內每個系統調用的耗時分析，可以用來分析web服務的響應時間都用在哪了：