Golang程序的性能分析heap memory profile和cpu prifile
把这个代码加载需要些heap memory profiler 日志的地方
import (
"runtime/pprof"
)
f, err := os.Create("HeapProfile.txt")
if err != nil {
log.Fatal(err)
}
pprof.WriteHeapProfile(f)
f.Close()
启动程序后， 运行到这个地方会生成 HeapProfile.txt 文件
go>go tool pprof
usage: pprof [options] [binary] <profile source> ...
D:\smsc\go>go tool pprof -text -inuse_objects smsc_cdr.exe HeapProfile.txt
386123 of 386124 total ( 100%)
Dropped 33 nodes (cum <= 1930)
flat flat% sum% cum cum%
229380 59.41% 59.41% 287809 74.54% main.generateCdr
76468 19.80% 79.21% 98314 25.46% github.com/Shopify/sarama.(*produceSet).add
38404 9.95% 89.16% 38404 9.95% bytes.makeSlice
21846 5.66% 94.81% 21846 5.66% github.com/Shopify/sarama.(*StringEncoder).Encode
16384 4.24% 99.06% 16384 4.24% github.com/lib/pq.textDecode
3641 0.94% 100% 3641 0.94% database/sql.convertAssign
0 0% 100% 38404 9.95% bytes.(*Buffer).WriteString
0 0% 100% 38404 9.95% bytes.(*Buffer).grow
0 0% 100% 16384 4.24% database/sql.(*Rows).Next
0 0% 100% 3641 0.94% database/sql.(*Rows).Scan
0 0% 100% 38404 9.95% encoding/json.(*encodeState).marshal
0 0% 100% 38404 9.95% encoding/json.(*encodeState).reflectValue
0 0% 100% 38404 9.95% encoding/json.(*encodeState).string
0 0% 100% 38404 9.95% encoding/json.(*structEncoder).(encoding/json.encode)-fm
0 0% 100% 38404 9.95% encoding/json.(*structEncoder).encode
0 0% 100% 38404 9.95% encoding/json.Marshal
0 0% 100% 38404 9.95% encoding/json.stringEncoder
0 0% 100% 98314 25.46% github.com/Shopify/sarama.(*brokerProducer).(github.com/Shopify/sarama.run)-fm
0 0% 100% 98314 25.46% github.com/Shopify/sarama.(*brokerProducer).run
0 0% 100% 98315 25.46% github.com/Shopify/sarama.withRecover
0 0% 100% 16384 4.24% github.com/lib/pq.(*rows).Next
0 0% 100% 16384 4.24% github.com/lib/pq.decode
0 0% 100% 287809 74.54% main.main
0 0% 100% 386124 100% runtime.goexit
go tool pprof
D:\smsc\go>go tool pprof smsc_cdr.exe HeapProfile.txt
Entering interactive mode (type "help" for commands)
(pprof) top
79.58MB of 79.58MB total ( 100%)
Dropped 28 nodes (cum <= 0.40MB)
Showing top 10 nodes out of 30 (cum >= 0.50MB)
flat flat% sum% cum cum%
51.57MB 64.80% 64.80% 51.57MB 64.80% bytes.makeSlice
14.85MB 18.66% 83.47% 14.85MB 18.66% github.com/Shopify/sarama.encode
5.66MB 7.11% 90.57% 6.66MB 8.36% github.com/Shopify/sarama.(*produceSet).add
5.50MB 6.91% 97.49% 58.07MB 72.97% main.generateCdr
1MB 1.26% 98.74% 1MB 1.26% github.com/Shopify/sarama.(*StringEncoder).Encode
0.50MB 0.63% 99.37% 0.50MB 0.63% database/sql.convertAssign
0.50MB 0.63% 100% 0.50MB 0.63% github.com/lib/pq.textDecode
0 0% 100% 51.57MB 64.80% bytes.(*Buffer).WriteString
0 0% 100% 51.57MB 64.80% bytes.(*Buffer).grow
0 0% 100% 0.50MB 0.63% database/sql.(*Rows).Next
(pprof) peek
79.58MB of 79.58MB total ( 100%)
Dropped 28 nodes (cum <= 0.40MB)
----------------------------------------------------------+-------------
flat flat% sum% cum cum% calls calls% + context
----------------------------------------------------------+-------------
51.57MB 100% | bytes.(*Buffer).grow
51.57MB 64.80% 64.80% 51.57MB 64.80% | bytes.makeSlice
----------------------------------------------------------+-------------
14.85MB 100% | github.com/Shopify/sarama.(*Broker).send
14.85MB 18.66% 83.47% 14.85MB 18.66% | github.com/Shopify/sarama.encode
----------------------------------------------------------+-------------
6.66MB 100% | github.com/Shopify/sarama.(*brokerProducer).run
5.66MB 7.11% 90.57% 6.66MB 8.36% | github.com/Shopify/sarama.(*produceSet).add
1MB 100% | github.com/Shopify/sarama.(*StringEncoder).Encode
----------------------------------------------------------+-------------
58.07MB 100% | main.main
5.50MB 6.91% 97.49% 58.07MB 72.97% | main.generateCdr
51.57MB 98.10% | encoding/json.Marshal
0.50MB 0.95% | database/sql.(*Rows).Scan
0.50MB 0.95% | database/sql.(*Rows).Next
----------------------------------------------------------+-------------
1MB 100% | github.com/Shopify/sarama.(*produceSet).add
1MB 1.26% 98.74% 1MB 1.26% | github.com/Shopify/sarama.(*StringEncoder).Encode
----------------------------------------------------------+-------------
0.50MB 100% | database/sql.(*Rows).Scan
0.50MB 0.63% 99.37% 0.50MB 0.63% | database/sql.convertAssign
----------------------------------------------------------+-------------
0.50MB 100% | github.com/lib/pq.decode
0.50MB 0.63% 100% 0.50MB 0.63% | github.com/lib/pq.textDecode
----------------------------------------------------------+-------------
51.57MB 100% | encoding/json.(*encodeState).string
0 0% 100% 51.57MB 64.80% | bytes.(*Buffer).WriteString
51.57MB 100% | bytes.(*Buffer).grow
----------------------------------------------------------+-------------
51.57MB 100% | bytes.(*Buffer).WriteString
0 0% 100% 51.57MB 64.80% | bytes.(*Buffer).grow
51.57MB 100% | bytes.makeSlice
通过http接口
http://docs.studygolang.com/pkg/net/http/pprof/
import "http"
import _ "net/http/pprof"
go func() {
log.Println(http.ListenAndServe("localhost:6060", nil))
}()
就可以通过下面这些命令，直接连接远程http接口来做性能分析了
go tool pprof http://localhost:6060/debug/pprof/heap
go tool pprof http://localhost:6060/debug/pprof/profile?seconds=30
wget http://localhost:6060/debug/pprof/trace?seconds=5
go tool pprof http://localhost:6060/debug/pprof/mutex
导出svg格式的函数调用树
(pprof) svg
Generating report in profile001.svg