[go]zap配合logrotate实现日志滚动

Posted 2022-11-29 adream307

tags:

篇首语：本文由小常识网(cha138.com)小编为大家整理，主要介绍了[go]zap配合logrotate实现日志滚动相关的知识，希望对你有一定的参考价值。

zap是Uber 提供的GoLang高性能日志库，zap 本身并不提供日志滚动功能，官方 FAQ 提到，可以使用Linux系统自带的 logrotate 或lumberjack实现日志滚动功能

lumberjack 只能向文件输出日志，如果我们希望同时向stderr 和文件输出日志，只能使用 logrotate 配合自定义 WriteSyncer 实现了

`Go`代码

package main

import (
	"bufio"
	"context"
	"fmt"
	"go.uber.org/zap"
	"go.uber.org/zap/zapcore"
	"os"
	"os/signal"
	"sync"
	"syscall"
)

type LogWriter struct 
	FileName string
	fp       *os.File
	mu       sync.Mutex
	fileCnt  int
	sig      chan os.Signal
	ctx      context.Context
	cancel   context.CancelFunc


func (m *LogWriter) Write(p []byte) (n int, err error) 
	m.mu.Lock()
	defer m.mu.Unlock()
	if m.fp == nil 
		if err := m.open(); err != nil 
			return 0, nil
		
	
	if m.fileCnt == 0 
		m.fileCnt++
		m.ctx, m.cancel = context.WithCancel(context.Background())
		m.sig = make(chan os.Signal, 1)
		signal.Notify(m.sig, syscall.SIGHUP)
		go func() 
			for 
				select 
				case <-m.ctx.Done():
					return
				case _, ok := <-m.sig:
					if ok == false 
						return
					
					func() 
						m.mu.Lock()
						defer m.mu.Unlock()
						if m.fp != nil 
							_ = m.fp.Close()
							m.fp = nil
						
					()
				
			
		()
	
	_, _ = os.Stderr.Write(p)
	return m.fp.Write(p)


func (m *LogWriter) Sync() error 
	m.mu.Lock()
	defer m.mu.Unlock()
	_ = os.Stderr.Sync()
	return m.fp.Sync()


func (m *LogWriter) Close() error 
	m.mu.Lock()
	defer m.mu.Unlock()
	m.cancel()
	if m.fp != nil 
		if err := m.fp.Close(); err != nil 
			return err
		
	
	if m.sig != nil 
		signal.Reset(syscall.SIGHUP)
		close(m.sig)
	
	return nil


func (m *LogWriter) open() error 
	if m.fp != nil 
		if err := m.fp.Sync(); err != nil 
			return err
		
		if err := m.fp.Close(); err != nil 
			return err
		
	
	fp, err := os.OpenFile(m.FileName, os.O_APPEND|os.O_WRONLY|os.O_CREATE, 0644)
	if err != nil 
		return err
	
	m.fp = fp
	return nil


func main() 
	zapid, _ := os.OpenFile("/tmp/zap.pid", os.O_CREATE|os.O_WRONLY, 0644)
	_, _ = zapid.WriteString(fmt.Sprintf("%d", os.Getpid()))
	_ = zapid.Sync()
	_ = syscall.Flock(int(zapid.Fd()), syscall.LOCK_EX)

	mylog := &LogWriterFileName: "/tmp/zap.log"
	cfg := zap.NewProductionEncoderConfig()
	cfg.EncodeTime = zapcore.ISO8601TimeEncoder
	core := zapcore.NewCore(zapcore.NewJSONEncoder(cfg), mylog, zap.InfoLevel)
	logger := zap.New(core)

	reader := bufio.NewReader(os.Stdin)
	for i := 0; i < 10; i++ 
		logger.Debug("debug msg", zap.Int("idx", i))
		logger.Info("info msg", zap.Int("idx", i))
		logger.Error("error msg", zap.Int("idx", i))
		logger.Warn("warn msg", zap.Int("idx", i))
		fmt.Printf("====================")
		_, _, _ = reader.ReadRune()
	
	_ = mylog.Close()

	_ = syscall.Flock(int(zapid.Fd()), syscall.LOCK_UN)
	_ = zapid.Close()

编译生成测试程序 zap_exp

go build zap_exp.go

`logrotate` 配置文件

/tmp/zap.log 
    su robin robin
    missingok
    daily
    minsize 1M 
    create 0644 robin robin
    rotate 12
    postrotate
        PIDFILE=/tmp/zap.pid
        if [ -f "$PIDFILE" ] ; then kill -HUP $(cat "$PIDFILE") ; fi
    endscript

zap.conf 的 owner 和 group 必须为 root，并且 file mode 为 0644 ，只有 root 可以读写，其它只读

su robin robin : 运行 zap_exp 程序的用户名及坐在的 Group，本测试中 user 和 group 均为 robin
create 0644 robin robin : 日志滚动后，重新生成日志文件的 file mode 及 ower 和 group
postrotate : 需要配合zap_exp.go 源码分析
- zap_exp.go 在 main 函数启动后首先生成 /tmp/zap.pid文件，然后将当前程序的 pid 写入该文件，并锁定该文件
- LogWriter 设置 signal.Notify(m.sig, syscall.SIGHUP) 表明 zap_exp 程序接受 HUP 信号
- LogWriter 收到 HUP 信号后关闭当前日志文件
- LogWriter写日志文件时，如果日志文件被关闭，程序自动重新创建
- logrotate 程序运行时，首先从/tmp/zap.pid文件读取zap_exp 程序的 pid ，然后向zap_exp 程序发送 HUP 信号

测试

运行 zap_exp 程序，分别在 idx:3 和 idx:6 时手动强制执行 logrotate 程序

logrotate -s ./log.status -vf ./zap.conf

zap_exp 运行结果

./zap_exp
"level":"info","ts":"2020-09-10T16:53:54.896+0800","msg":"info msg","idx":0
"level":"error","ts":"2020-09-10T16:53:54.896+0800","msg":"error msg","idx":0
"level":"warn","ts":"2020-09-10T16:53:54.896+0800","msg":"warn msg","idx":0
====================
"level":"info","ts":"2020-09-10T16:53:56.628+0800","msg":"info msg","idx":1
"level":"error","ts":"2020-09-10T16:53:56.628+0800","msg":"error msg","idx":1
"level":"warn","ts":"2020-09-10T16:53:56.628+0800","msg":"warn msg","idx":1
====================
"level":"info","ts":"2020-09-10T16:53:57.060+0800","msg":"info msg","idx":2
"level":"error","ts":"2020-09-10T16:53:57.060+0800","msg":"error msg","idx":2
"level":"warn","ts":"2020-09-10T16:53:57.060+0800","msg":"warn msg","idx":2
====================
"level":"info","ts":"2020-09-10T16:53:57.393+0800","msg":"info msg","idx":3
"level":"error","ts":"2020-09-10T16:53:57.393+0800","msg":"error msg","idx":3
"level":"warn","ts":"2020-09-10T16:53:57.393+0800","msg":"warn msg","idx":3
====================
"level":"info","ts":"2020-09-10T16:54:31.706+0800","msg":"info msg","idx":4
"level":"error","ts":"2020-09-10T16:54:31.706+0800","msg":"error msg","idx":4
"level":"warn","ts":"2020-09-10T16:54:31.706+0800","msg":"warn msg","idx":4
====================
"level":"info","ts":"2020-09-10T16:54:32.050+0800","msg":"info msg","idx":5
"level":"error","ts":"2020-09-10T16:54:32.050+0800","msg":"error msg","idx":5
"level":"warn","ts":"2020-09-10T16:54:32.050+0800","msg":"warn msg","idx":5
====================
"level":"info","ts":"2020-09-10T16:54:32.359+0800","msg":"info msg","idx":6
"level":"error","ts":"2020-09-10T16:54:32.359+0800","msg":"error msg","idx":6
"level":"warn","ts":"2020-09-10T16:54:32.359+0800","msg":"warn msg","idx":6
====================
"level":"info","ts":"2020-09-10T16:54:36.955+0800","msg":"info msg","idx":7
"level":"error","ts":"2020-09-10T16:54:36.956+0800","msg":"error msg","idx":7
"level":"warn","ts":"2020-09-10T16:54:36.956+0800","msg":"warn msg","idx":7
====================
"level":"info","ts":"2020-09-10T16:54:37.255+0800","msg":"info msg","idx":8
"level":"error","ts":"2020-09-10T16:54:37.255+0800","msg":"error msg","idx":8
"level":"warn","ts":"2020-09-10T16:54:37.255+0800","msg":"warn msg","idx":8
====================
"level":"info","ts":"2020-09-10T16:54:37.712+0800","msg":"info msg","idx":9
"level":"error","ts":"2020-09-10T16:54:37.712+0800","msg":"error msg","idx":9
"level":"warn","ts":"2020-09-10T16:54:37.712+0800","msg":"warn msg","idx":9
====================

在 /tmp 可以找到如下3个文件 zap.log zap.log.1 zap.log.2
zap.log.2 内容如下，刚好对应 idx:3 时执行 logrotate 时产生的日志滚动

"level":"info","ts":"2020-09-10T16:53:54.896+0800","msg":"info msg","idx":0
"level":"error","ts":"2020-09-10T16:53:54.896+0800","msg":"error msg","idx":0
"level":"warn","ts":"2020-09-10T16:53:54.896+0800","msg":"warn msg","idx":0
"level":"info","ts":"2020-09-10T16:53:56.628+0800","msg":"info msg","idx":1
"level":"error","ts":"2020-09-10T16:53:56.628+0800","msg":"error msg","idx":1
"level":"warn","ts":"2020-09-10T16:53:56.628+0800","msg":"warn msg","idx":1
"level":"info","ts":"2020-09-10T16:53:57.060+0800","msg":"info msg","idx":2
"level":"error","ts":"2020-09-10T16:53:57.060+0800","msg":"error msg","idx":2
"level":"warn","ts":"2020-09-10T16:53:57.060+0800","msg":"warn msg","idx":2
"level":"info","ts":"2020-09-10T16:53:57.393+0800","msg":"info msg","idx":3
"level":"error","ts":"2020-09-10T16:53:57.393+0800","msg":"error msg","idx":3
"level":"warn","ts":"2020-09-10T16:53:57.393+0800","msg":"warn msg","idx":3

参考资料

以上是关于[go]zap配合logrotate实现日志滚动的主要内容，如果未能解决你的问题，请参考以下文章

高性能 Go 日志库 zap 设计与实现

Go语言项目中使用zap日志库（翻译）

如何使用 uber-go/zap 根据日志级别记录到标准输出或标准错误？

Go语言系列之日志库zap

[go]zap配合logrotate实现日志滚动

Go代码

logrotate 配置文件

测试

参考资料

`Go`代码

`logrotate` 配置文件