一说到写日志,大家可能推荐一堆的开源日志框架,如:Log4Net、NLog,这些日志框架确实也不错,比较强大也比较灵活,但也正因为又强大又灵活,导致我们使用他们时需要引用一些DLL,同时还要学习各种用法及配置文件,这对于有些小工具、小程序、小网站来说,有点“杀鸡焉俺用牛刀”的感觉,而且如果对这些日志框架不了解,可能输出来的日志性能或效果未毕是与自己所想的,鉴于这几个原因,我自己重复造轮子,编写了一个轻量级的异步写日志的实用工具类(LogAsyncWriter),这个类还是比较简单的,实现思路也很简单,就是把消息日志先入内存队列,然后由异步监听线程从队列中取出日志并批量输出到本地文件中,同时参照各大日志框架,对单个文件过大采取分割生成多个日志文件。
经测试发现性能非常不错,先看示例使用代码:(采取并发多线程同时写入1000000万条日志)
Task.Factory.StartNew(() => { DateTime startTime = DateTime.Now; int logCount = 1000000; Parallel.For(1, logCount, (i) => { if (i % 2 == 0) { LogAsyncWriter.Default.Error("测试并发写错误日志-" + i.ToString(), "TestClass.TestLog", i.ToString()); } else { LogAsyncWriter.Default.Info("测试并发写普通日志-" + i.ToString(), "TestClass.TestLog", i.ToString()); } }); this.Invoke(new MethodInvoker(() => { MessageBox.Show(DateTime.Now.ToString() + "," + logCount + "条日志写完了!,耗时:" + (DateTime.Now - startTime).TotalMilliseconds + "ms"); })); }); MessageBox.Show(DateTime.Now.ToString() + ",同步方法已结束"); }
执行效果如下图示:
因为采用异步,故方法先走到结尾,输出了同步的MsgBox,随后弹出的是100W日志输出到文件后的耗时MsgBox,从截图可以看出,不足1S(当然这里的1S不是真实的输出到本地方件,而是把所有的日志推到了Queue中而矣,但不影响不阻塞业务处理),而本地日志文件的大小达到了263MB(设置最大值的MaxSizeBackup,使其不滚动备份),由此看性能是不错的;
因为是异步延迟输出到本地日志文件,故大家在代码中任意地方,比如:循环中都可以使用它,不用担心会影响你的正常的业务逻辑的执行效率;
如果采取滚动备份(设置MaxSizeBackup为一个合理值,默认为10MB),则生成的日志文件形式如下:(超过最大容量则生成同名文件+2位序号),超过一个月则会自动清除
打开日志文件会看到所有的日志内容:
有些人可能要问,输出格式是固定的吗?能否自定义布局,告诉你是可以的,我考虑到每个人对日志输出的格式都有严格的要求,故可以通过设置:LineLayoutRenderFormat属性来设置每条日志输出的格式内容
好了,介绍了使用功能及效果、性能,下面贴出源代码:
using System; using System.Collections.Concurrent; using System.Collections.Generic; using System.IO; using System.Linq; using System.Text; using System.Threading; using System.Threading.Tasks; using System.Text.RegularExpressions; namespace Zuowj.Common { /// <summary> /// 日志异步生成器 /// Author:Zuowenjun(http://www.zuowenjun.cn) /// Date:2018-6-14 /// </summary> public class LogAsyncWriter { public const string InfoLevel = "INFO"; public const string WarnLevel = "WARN"; public const string ErrorLevel = "ERROR"; private readonly ConcurrentQueue<string[]> logMsgQueue = new ConcurrentQueue<string[]>(); private readonly CancellationTokenSource cts = null; private string lineLayoutRenderFormat = "[{0:yyyy-MM-dd HH:mm:ss}] {1} {2} {3}:{4},Trace:{5};Other1:{6},Other2:{7},Other3:{8}"; private long maxSizeBackup = 10485760L;//默认10MB private string todayLogName = null; private static readonly LogAsyncWriter instance = new LogAsyncWriter(); private LogAsyncWriter() { cts = new CancellationTokenSource(); ListenSaveLogAsync(cts.Token); } private void ListenSaveLogAsync(CancellationToken cancellationToken) { Task.Factory.StartNew(() => { DateTime lastSaveLogTime = DateTime.Now; while (!cancellationToken.IsCancellationRequested)//如果没有取消线程,则一直监听执行写LOG { if (logMsgQueue.Count >= 10 || (logMsgQueue.Count > 0 && (DateTime.Now - lastSaveLogTime).TotalSeconds > 30))//如是待写日志消息累计>=10条或上一次距离现在写日志时间超过30s则需要批量提交日志 { List<string[]> logMsgList = new List<string[]>(); string[] logMsgItems = null; while (logMsgList.Count < 10 && logMsgQueue.TryDequeue(out logMsgItems)) { logMsgList.Add(logMsgItems); } WriteLog(logMsgList); lastSaveLogTime = DateTime.Now; } else { SpinWait.SpinUntil(() => logMsgQueue.Count >= 10, 5000);//自旋等待直到日志队列有>=10的记录或超时5S后再进入下一轮的判断 } } }, cancellationToken); } private string GetLogFilePath() { string logFileDir = Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "Logs"); if (!Directory.Exists(logFileDir)) { Directory.CreateDirectory(logFileDir); } string logDateStr = DateTime.Now.ToString("yyyyMMdd"); string logName = logDateStr; if (!string.IsNullOrEmpty(todayLogName) && todayLogName.StartsWith(logName)) { logName = todayLogName; } else { todayLogName = logName; } string logFilePath = Path.Combine(logFileDir, logName + ".log"); if (File.Exists(logFilePath)) { File.SetAttributes(logFilePath, FileAttributes.Normal); if (File.GetLastWriteTime(logFilePath).Month != DateTime.Today.Month) //30天滚动(删除旧的文件),防止日志文件过多 { File.Delete(logFilePath); string[] oldLogFiles = Directory.GetFiles(logFileDir, string.Format("{0}-##.log", logDateStr), SearchOption.TopDirectoryOnly); foreach (string fileName in oldLogFiles) { File.SetAttributes(fileName, FileAttributes.Normal); File.Delete(fileName); } } else if (new FileInfo(logFilePath).Length > MaxSizeBackup) { Regex rgx = new Regex(@"^d{8}-(?<fnum>d{2})$"); int fnum = 2; if (rgx.IsMatch(logName)) { fnum = int.Parse(rgx.Match(logName).Groups["fnum"].Value) + 1; } logName = string.Format("{0}-{1:D2}", logDateStr, fnum); todayLogName = logName; logFilePath = Path.Combine(logFileDir, logName + ".log"); } } return logFilePath; } private void WriteLog(IEnumerable<string[]> logMsgs) { try { List<string> logMsgLines = new List<string>(); foreach (var logMsgItems in logMsgs) { var logMsgLineFields = (new object[] { DateTime.Now }).Concat(logMsgItems).ToArray(); string logMsgLineText = string.Format(LineLayoutRenderFormat, logMsgLineFields); logMsgLines.Add(logMsgLineText); } string logFilePath = GetLogFilePath(); File.AppendAllLines(logFilePath, logMsgLines); } catch { } } public static LogAsyncWriter Default { get { return instance; } } public string LineLayoutRenderFormat { get { return lineLayoutRenderFormat; } set { if (string.IsNullOrWhiteSpace(value)) { throw new ArgumentException("无效的LineLayoutRenderFormat属性值"); } lineLayoutRenderFormat = value; } } public long MaxSizeBackup { get { return maxSizeBackup; } set { if (value <= 0) { throw new ArgumentException("无效的MaxSizeBackup属性值"); } maxSizeBackup = value; } } public void SaveLog(string logLevel, string msg, string source, string detailTrace = null, string other1 = null, string other2 = null, string other3 = null) { logMsgQueue.Enqueue(new[] { logLevel, Thread.CurrentThread.ManagedThreadId.ToString(), source, msg, detailTrace ?? string.Empty, other1 ?? string.Empty, other2 ?? string.Empty, other3 ?? string.Empty }); } public void Info(string msg, string source, string detailTrace = null, string other1 = null, string other2 = null, string other3 = null) { SaveLog(InfoLevel, msg, source, detailTrace, other1, other2, other3); } public void Warn(string msg, string source, string detailTrace = null, string other1 = null, string other2 = null, string other3 = null) { SaveLog(WarnLevel, msg, source, detailTrace, other1, other2, other3); } public void Error(string msg, string source, string detailTrace = null, string other1 = null, string other2 = null, string other3 = null) { SaveLog(ErrorLevel, msg, source, detailTrace, other1, other2, other3); } public void Error(Exception ex, string source, string other1 = null, string other2 = null, string other3 = null) { SaveLog(ErrorLevel, ex.Message, source, ex.StackTrace, other1, other2, other3); } ~LogAsyncWriter() { cts.Cancel(); } } }
代码重点说明:
1.各种日志方法入参解释:
i. Msg:日志消息(一般是指简要消息)
ii. Source:日志产生源(一般是指该日志是由哪个位置产生的,可以定义为:类.方法名)
iii. detailTrace:日志详情(一般是指堆栈信息或日志更具体的信息)
iv. other1, other2, other3:备用日志字段,可根据实际情况记录相关信息,比如:入参、返参,执行耗时等;
v. log Level:日志消息级别一般有很多级别,但常用的只有3类,即:Info=普通日志消息,Warn=警告日志消息,Error=错误日志消息;
2.核心异步批量写日志的方法:(采用后台线程监听日志消息队列,当达到10条日志消息或写日志的时间间隔超过1分钟,则会批量提交1次日志,解决了普通的同步写日志方法造成写压力过大,且存在阻塞业务逻辑的情况)
3.采用单例模式,当第一次使用该类时,则会启动异步监听线程,当该类释放时(一般指应用进程关闭时,会发送通知线程取消监听,避免一切可能的线程驻留问题)
好了就介绍到这里,大家若想试用或想改造,可以直接复制上述代码,不足之处可以指出,谢谢!