一般作为服务端的应用,必须要有相应的日志,否则问题怎么排查呢?
而日志怎么打印,也是一个技术活。不然java中也不会存在N多厂商争相提供日志框架了!
而日志滚动则往往也是刚需,毕竟没人能保证日志的量及可阅读性。日志滚动实现主要有两个大方向:
1. 让应用服务自行打印,打印到时也完全由应用决定!
2. 借助第三方的工具进行日志打印,这种一般要借助于控制台或者agent!
3. 让日志框架提供日志滚动功能,自行管理日志;这样做有个好处就是,应用自带,无需外部处理。坏处就是要完全依赖该应用,会影响该应用的性能,且如果该应用存在bug,则功能就不敢保证了。(稍后我会以logback的日志滚动说明)
4. 借助第三方的工具进行日志滚动;这样做的好处是滚动功能更独立,对代码无入侵,即使真的有问题,大不了把它干掉也没关系;另外,第三方工具不会因为应用本身的bug而导致滚动异常,从而保证了有足够的排查依据。(稍后我会以cronolog进行讲解滚动实现);
具体日志滚动实现
1. 使用应用打印的方式:如logback的rollingpolicy,则自带滚动日志功能!但是坑多!
1.1. 首先我们看下日志滚动的配置:(在 logback.xml 配置)
<!--输出到文件--> <appender name="file" class="ch.qos.logback.core.rolling.RollingFileAppender"> <file>${log_path}/api.ln.log</file> <rollingPolicy class="ch.qos.logback.core.rolling.TimeBasedRollingPolicy" > <fileNamePattern>${log_path}/api.%d{yyyy-MM-dd_HH}.log</fileNamePattern> <!-- keep 10 days' worth of history capped at 8GB total size --> <maxHistory>10</maxHistory> <totalSizeCap>8GB</totalSizeCap> </rollingPolicy> <encoder> <pattern>%d{MM-dd HH:mm:ss.SSS} [%thread] %-5level %logger{36} - %msg%n</pattern> </encoder> </appender>
这里配置以时间为滚动标准,每小时滚动一次!最大保留10天日志,总共大小不超过8G。我们后面来看下他的效果!
1.2. 看下滚动代码!
首先,日志滚动会有相应的线程一直在跑(不管是应用实现还是第三方实现都是这样,否则怎么随时检测滚动时机呢)!
在 EventPlayer中,有个play方法,此时会决断是否是 EndEvent, 如果是的话就会调用后台线程生成!
// ch.qos.logback.core.joran.spi.EventPlayer public void play(List<SaxEvent> aSaxEventList) { eventList = aSaxEventList; SaxEvent se; for (currentIndex = 0; currentIndex < eventList.size(); currentIndex++) { se = eventList.get(currentIndex); if (se instanceof StartEvent) { interpreter.startElement((StartEvent) se); // invoke fireInPlay after startElement processing interpreter.getInterpretationContext().fireInPlay(se); } if (se instanceof BodyEvent) { // invoke fireInPlay before characters processing interpreter.getInterpretationContext().fireInPlay(se); interpreter.characters((BodyEvent) se); } // rollingPollicy 在此处调唤醒 if (se instanceof EndEvent) { // invoke fireInPlay before endElement processing interpreter.getInterpretationContext().fireInPlay(se); interpreter.endElement((EndEvent) se); } } }
然后,几经转换,就到了Interpreter 了,这里会做一个死循环,一直在监听!
// ch.qos.logback.core.joran.spi.Interpreter private void callEndAction(List<Action> applicableActionList, String tagName) { if (applicableActionList == null) { return; } // logger.debug("About to call end actions on node: [" + localName + "]"); Iterator<Action> i = applicableActionList.iterator(); while (i.hasNext()) { Action action = i.next(); // now let us invoke the end method of the action. We catch and report // any eventual exceptions try { action.end(interpretationContext, tagName); } catch (ActionException ae) { // at this point endAction, there is no point in skipping children as // they have been already processed cai.addError("ActionException in Action for tag [" + tagName + "]", ae); } catch (RuntimeException e) { // no point in setting skip cai.addError("RuntimeException in Action for tag [" + tagName + "]", e); } } }
最后,就会调用 RollingPolicy 的start()了,这里是 TimeBasedRollingPollicy .
// ch.qos.logback.core.rolling.TimeBasedRollingPolicy public void start() { // set the LR for our utility object renameUtil.setContext(this.context); // find out period from the filename pattern if (fileNamePatternStr != null) { fileNamePattern = new FileNamePattern(fileNamePatternStr, this.context); determineCompressionMode(); } else { addWarn(FNP_NOT_SET); addWarn(CoreConstants.SEE_FNP_NOT_SET); throw new IllegalStateException(FNP_NOT_SET + CoreConstants.SEE_FNP_NOT_SET); } compressor = new Compressor(compressionMode); compressor.setContext(context); // wcs : without compression suffix fileNamePatternWithoutCompSuffix = new FileNamePattern(Compressor.computeFileNameStrWithoutCompSuffix(fileNamePatternStr, compressionMode), this.context); addInfo("Will use the pattern " + fileNamePatternWithoutCompSuffix + " for the active file"); if (compressionMode == CompressionMode.ZIP) { String zipEntryFileNamePatternStr = transformFileNamePattern2ZipEntry(fileNamePatternStr); zipEntryFileNamePattern = new FileNamePattern(zipEntryFileNamePatternStr, context); } // 默认会使用 DefaultTimeBasedFileNamingAndTriggeringPolicy 进行滚动 if (timeBasedFileNamingAndTriggeringPolicy == null) { timeBasedFileNamingAndTriggeringPolicy = new DefaultTimeBasedFileNamingAndTriggeringPolicy<E>(); } timeBasedFileNamingAndTriggeringPolicy.setContext(context); timeBasedFileNamingAndTriggeringPolicy.setTimeBasedRollingPolicy(this); timeBasedFileNamingAndTriggeringPolicy.start(); if (!timeBasedFileNamingAndTriggeringPolicy.isStarted()) { addWarn("Subcomponent did not start. TimeBasedRollingPolicy will not start."); return; } // the maxHistory property is given to TimeBasedRollingPolicy instead of to // the TimeBasedFileNamingAndTriggeringPolicy. This makes it more convenient // for the user at the cost of inconsistency here. if (maxHistory != UNBOUND_HISTORY) { archiveRemover = timeBasedFileNamingAndTriggeringPolicy.getArchiveRemover(); archiveRemover.setMaxHistory(maxHistory); archiveRemover.setTotalSizeCap(totalSizeCap.getSize()); if (cleanHistoryOnStart) { addInfo("Cleaning on start up"); Date now = new Date(timeBasedFileNamingAndTriggeringPolicy.getCurrentTime()); cleanUpFuture = archiveRemover.cleanAsynchronously(now); } } else if (!isUnboundedTotalSizeCap()) { addWarn("'maxHistory' is not set, ignoring 'totalSizeCap' option with value ["+totalSizeCap+"]"); } // 调用父类start(), 设置启动标识,不允许多次调用初始化 super.start(); } // DefaultTimeBasedFileNamingAndTriggeringPolicy 的实现,设置类功能主要还是调用 TimeBasedFileNamingAndTriggeringPolicy 的方法,而其自身,则是处理一些异常情况,以及开启一个 Remover, 供具体的实现调用 @Override public void start() { super.start(); if (!super.isErrorFree()) return; if(tbrp.fileNamePattern.hasIntegerTokenCOnverter()) { addError("Filename pattern ["+tbrp.fileNamePattern+"] contains an integer token converter, i.e. %i, INCOMPATIBLE with this configuration. Remove it."); return; } archiveRemover = new TimeBasedArchiveRemover(tbrp.fileNamePattern, rc); archiveRemover.setContext(context); started = true; } // TimeBasedFileNamingAndTriggeringPolicy, 则实际处理日志的滚动逻辑了 public void start() { DateTokenConverter<Object> dtc = tbrp.fileNamePattern.getPrimaryDateTokenConverter(); if (dtc == null) { throw new IllegalStateException("FileNamePattern [" + tbrp.fileNamePattern.getPattern() + "] does not contain a valid DateToken"); } if (dtc.getTimeZone() != null) { rc = new RollingCalendar(dtc.getDatePattern(), dtc.getTimeZone(), Locale.getDefault()); } else { rc = new RollingCalendar(dtc.getDatePattern()); } addInfo("The date pattern is '" + dtc.getDatePattern() + "' from file name pattern '" + tbrp.fileNamePattern.getPattern() + "'."); rc.printPeriodicity(this); if (!rc.isCollisionFree()) { addError("The date format in FileNamePattern will result in collisions in the names of archived log files."); addError(CoreConstants.MORE_INFO_PREFIX + COLLIDING_DATE_FORMAT_URL); withErrors(); return; } setDateInCurrentPeriod(new Date(getCurrentTime())); if (tbrp.getParentsRawFileProperty() != null) { File currentFile = new File(tbrp.getParentsRawFileProperty()); if (currentFile.exists() && currentFile.canRead()) { setDateInCurrentPeriod(new Date(currentFile.lastModified())); } } addInfo("Setting initial period to " + dateInCurrentPeriod); computeNextCheck(); }
经过如上初始化动作之后,发现并没有启动相应的轮循线程,所以这个点也是超出简单的认知了,不管怎么样,我们还要继续的!我们先来看一下 RollingFileAppender 的 append() 逻辑吧,毕竟它才是log的接入口!
// ch.qos.logback.core.ch.qos.logback.core.rolling.RollingFileAppender, 其接入口为: UnsynchronizedAppenderBase.doAppend() // ch.qos.logback.core.OutputStreamAppender @Override protected void append(E eventObject) { if (!isStarted()) { return; } // 调用 RollingFileAppender 实现 subAppend(eventObject); } // ch.qos.logback.core.ch.qos.logback.core.rolling.RollingFileAppender @Override protected void subAppend(E event) { // The roll-over check must precede actual writing. This is the // only correct behavior for time driven triggers. // We need to synchronize on triggeringPolicy so that only one rollover // occurs at a time synchronized (triggeringPolicy) { if (triggeringPolicy.isTriggeringEvent(currentlyActiveFile, event)) { rollover(); } } super.subAppend(event); }
其中,rollover()就是其滚动逻辑!
所以,看到了吧!这里的文件滚动,是依赖于外部写入的,原因是为了写入的线程安全,保证文件的完整性!
换句话说就是,如果在滚动的这个时机,如果有外部写入,那么,文件得以滚动,否则,不会主动滚动文件!如果外部一直没日志写入,就不会存在日志滚动!
我们先来看下滚动的条件吧: triggeringPolicy.isTriggeringEvent(currentlyActiveFile, event)
// ch.qos.logback.core.rolling.DefaultTimeBasedFileNamingAndTriggeringPolicy public boolean isTriggeringEvent(File activeFile, final E event) { long time = getCurrentTime(); if (time >= nextCheck) { Date dateOfElapsedPeriod = dateInCurrentPeriod; addInfo("Elapsed period: " + dateOfElapsedPeriod); elapsedPeriodsFileName = tbrp.fileNamePatternWithoutCompSuffix.convert(dateOfElapsedPeriod); setDateInCurrentPeriod(time); computeNextCheck(); return true; } else { return false; } }
如上判断,即将当前时间与需要滚动的时间做对,大于滚动时间则返回 true, 并计算出下次需要滚动的时间,备用!
接下来,我们看下,具体的文件滚动实现!两个主逻辑: 1. 将文件更名滚动; 2. 重新创建一个新的目标文件,以使后续可以写入!
/** * Implemented by delegating most of the rollover work to a rolling policy. */ public void rollover() { // 此处lock为 ReentrantLock, 即是互斥锁,只能一个线程可访问! lock.lock(); try { // Note: This method needs to be synchronized because it needs exclusive // access while it closes and then re-opens the target file. // // make sure to close the hereto active log file! Renaming under windows // does not work for open files. this.closeOutputStream(); attemptRollover(); attemptOpenFile(); } finally { lock.unlock(); } } // 滚动文件逻辑,调用设置的 policy 实现进行滚动,此处我设置的是 TimeBasedRollingPolicy private void attemptRollover() { try { rollingPolicy.rollover(); } catch (RolloverFailure rf) { addWarn("RolloverFailure occurred. Deferring roll-over."); // we failed to roll-over, let us not truncate and risk data loss this.append = true; } } // ch.qos.logback.core.rolling.TimeBasedRollingPolicy rollover public void rollover() throws RolloverFailure { // when rollover is called the elapsed period's file has // been already closed. This is a working assumption of this method. String elapsedPeriodsFileName = timeBasedFileNamingAndTriggeringPolicy.getElapsedPeriodsFileName(); String elapsedPeriodStem = FileFilterUtil.afterLastSlash(elapsedPeriodsFileName); if (compressionMode == CompressionMode.NONE) { if (getParentsRawFileProperty() != null) { renameUtil.rename(getParentsRawFileProperty(), elapsedPeriodsFileName); } // else { nothing to do if CompressionMode == NONE and parentsRawFileProperty == null } } else { if (getParentsRawFileProperty() == null) { compressionFuture = compressor.asyncCompress(elapsedPeriodsFileName, elapsedPeriodsFileName, elapsedPeriodStem); } else { compressionFuture = renameRawAndAsyncCompress(elapsedPeriodsFileName, elapsedPeriodStem); } } if (archiveRemover != null) { Date now = new Date(timeBasedFileNamingAndTriggeringPolicy.getCurrentTime()); this.cleanUpFuture = archiveRemover.cleanAsynchronously(now); } }
TimeBasedRollingPolicy 的滚动方式为,重命名文件即可!即先获取外部设置的主写文件,然后根据新文件命名规则,生成一个新路径,然后重命名文件!重命名也是有些讲究的,有兴趣的同学可以查看下其重命名的实现!
// ch.qos.logback.core.rolling.helper.RenameUtil /** * A relatively robust file renaming method which in case of failure due to * src and target being on different volumes, falls back onto * renaming by copying. * * @param src * @param target * @throws RolloverFailure */ public void rename(String src, String target) throws RolloverFailure { if (src.equals(target)) { addWarn("Source and target files are the same [" + src + "]. Skipping."); return; } File srcFile = new File(src); if (srcFile.exists()) { // 如果目录不存在,会先去创建目录,所以你可以滚动到其他地方,而目录位置则不用管(权限除外) File targetFile = new File(target); createMissingTargetDirsIfNecessary(targetFile); addInfo("Renaming file [" + srcFile + "] to [" + targetFile + "]"); boolean result = srcFile.renameTo(targetFile); // 对于直接重命名失败,则会再次尝试,如果在不同的分区,则会使用一次文件复制的方式进行一次重命名,具体做法是,先把文件copy到新地址,然后再将当前文件删除 if (!result) { addWarn("Failed to rename file [" + srcFile + "] as [" + targetFile + "]."); Boolean areOnDifferentVolumes = areOnDifferentVolumes(srcFile, targetFile); if (Boolean.TRUE.equals(areOnDifferentVolumes)) { addWarn("Detected different file systems for source [" + src + "] and target [" + target + "]. Attempting rename by copying."); renameByCopying(src, target); return; } else { addWarn("Please consider leaving the [file] option of " + RollingFileAppender.class.getSimpleName() + " empty."); addWarn("See also " + RENAMING_ERROR_URL); } } } else { throw new RolloverFailure("File [" + src + "] does not exist."); } }
在做完日志重命名的滚动后,还有一个可能的工作,就是删除过期的日志!这个工作由 archiveRemover 来做,即之前在 DefaultTimeBasedFileNamingAndTriggeringPolicy 中创建的实例! 会调用其 archiveRemover.cleanAsynchronously(now);
public Future<?> cleanAsynchronously(Date now) { ArhiveRemoverRunnable runnable = new ArhiveRemoverRunnable(now); ExecutorService executorService = context.getScheduledExecutorService(); Future<?> future = executorService.submit(runnable); return future; }
在做删除过期日志时,会先获取一个 ExecutorService, 进行异步删除, 而这个 ExecutorService 默认开启 8 常驻线程,进行日志处理!
删除动作进行异步执行,从而避免影响业务执行!清理过程如下:
public class ArhiveRemoverRunnable implements Runnable { Date now; ArhiveRemoverRunnable(Date now) { this.now = now; } @Override public void run() { // 先清除当前文件,再根据设置的最大值,删除列表 clean(now); if (totalSizeCap != UNBOUNDED_TOTAL_SIZE_CAP && totalSizeCap > 0) { capTotalSize(now); } } } public void clean(Date now) { long nowInMillis = now.getTime(); // for a live appender periodsElapsed is expected to be 1 int periodsElapsed = computeElapsedPeriodsSinceLastClean(nowInMillis); lastHeartBeat = nowInMillis; if (periodsElapsed > 1) { addInfo("Multiple periods, i.e. " + periodsElapsed + " periods, seem to have elapsed. This is expected at application start."); } for (int i = 0; i < periodsElapsed; i++) { // 此处会根据 maxHistory 进行 -1 后清除文件,即: 只会清理 periodsElapsed 次历史日志 int offset = getPeriodOffsetForDeletionTarget() - i; Date dateOfPeriodToClean = rc.getEndOfNextNthPeriod(now, offset); cleanPeriod(dateOfPeriodToClean); } } public void cleanPeriod(Date dateOfPeriodToClean) { // 获取需要删除的文件列表,然后依次删除,如果文件夹内的文件全部被删除,则将文件夹删除 File[] matchingFileArray = getFilesInPeriod(dateOfPeriodToClean); for (File f : matchingFileArray) { addInfo("deleting " + f); f.delete(); } if (parentClean && matchingFileArray.length > 0) { File parentDir = getParentDir(matchingFileArray[0]); removeFolderIfEmpty(parentDir); } } // 按规则匹配需要删除的文件 protected File[] getFilesInPeriod(Date dateOfPeriodToClean) { String filenameToDelete = fileNamePattern.convert(dateOfPeriodToClean); File file2Delete = new File(filenameToDelete); if (fileExistsAndIsFile(file2Delete)) { return new File[] { file2Delete }; } else { return new File[0]; } } // 清理历史文件逻辑,注意要想清理历史文件,就一定要设置好 totalSizeCap, 否则,不会进行自动清理! void capTotalSize(Date now) { long totalSize = 0; long totalRemoved = 0; for (int offset = 0; offset < maxHistory; offset++) { Date date = rc.getEndOfNextNthPeriod(now, -offset); File[] matchingFileArray = getFilesInPeriod(date); descendingSortByLastModified(matchingFileArray); for (File f : matchingFileArray) { long size = f.length(); if (totalSize + size > totalSizeCap) { addInfo("Deleting [" + f + "]" + " of size " + new FileSize(size)); totalRemoved += size; f.delete(); } totalSize += size; } } addInfo("Removed " + new FileSize(totalRemoved) + " of files"); }
以上就是一个删除过期日志的逻辑,主要有几个点:
1. 只会进行清理 maxHistory 个周期的日志,即只会倒推 n 个周期内的日志;
2. 只会清理文件大小大于 totalSizeCap 大小以后的文件;(这个文件强依赖文件列表的排序,这里的排序是根据最后修改时间来排的)
3. maxHistory 并非最大保留天数,不要相信坑货文档,它只是一个扫描周期而已,不过这个值在上一步清理时会处理一次!
还有个细节,咱们得再来看看:滚动时机,按天,按小时,按分钟?
// 滚动时机判定 // ch.qos.logback.core.rolling.helper.RollingCalendar public Date getEndOfNextNthPeriod(Date now, int periods) { return innerGetEndOfNextNthPeriod(this, this.periodicityType, now, periods); } static private Date innerGetEndOfNextNthPeriod(Calendar cal, PeriodicityType periodicityType, Date now, int numPeriods) { cal.setTime(now); switch (periodicityType) { case TOP_OF_MILLISECOND: cal.add(Calendar.MILLISECOND, numPeriods); break; case TOP_OF_SECOND: cal.set(Calendar.MILLISECOND, 0); cal.add(Calendar.SECOND, numPeriods); break; case TOP_OF_MINUTE: cal.set(Calendar.SECOND, 0); cal.set(Calendar.MILLISECOND, 0); cal.add(Calendar.MINUTE, numPeriods); break; case TOP_OF_HOUR: cal.set(Calendar.MINUTE, 0); cal.set(Calendar.SECOND, 0); cal.set(Calendar.MILLISECOND, 0); cal.add(Calendar.HOUR_OF_DAY, numPeriods); break; case TOP_OF_DAY: cal.set(Calendar.HOUR_OF_DAY, 0); cal.set(Calendar.MINUTE, 0); cal.set(Calendar.SECOND, 0); cal.set(Calendar.MILLISECOND, 0); cal.add(Calendar.DATE, numPeriods); break; case TOP_OF_WEEK: cal.set(Calendar.DAY_OF_WEEK, cal.getFirstDayOfWeek()); cal.set(Calendar.HOUR_OF_DAY, 0); cal.set(Calendar.MINUTE, 0); cal.set(Calendar.SECOND, 0); cal.set(Calendar.MILLISECOND, 0); cal.add(Calendar.WEEK_OF_YEAR, numPeriods); break; case TOP_OF_MONTH: cal.set(Calendar.DATE, 1); cal.set(Calendar.HOUR_OF_DAY, 0); cal.set(Calendar.MINUTE, 0); cal.set(Calendar.SECOND, 0); cal.set(Calendar.MILLISECOND, 0); cal.add(Calendar.MONTH, numPeriods); break; default: throw new IllegalStateException("Unknown periodicity type."); } return cal.getTime(); }
可以看到其滚动的粒度: TOP_OF_MILLISECOND/TOP_OF_SECOND/TOP_OF_MINUTE/TOP_OF_HOUR/TOP_OF_DAY/TOP_OF_WEEK/TOP_OF_MONTH, 要说起来,粒度还是很细的哦!至于能不能真的有用,另说了!
总结下logback的滚动方式!
1. 在写入的时机进行滚动时机检查,合适则进行滚动;
2. 同步滚动操作,保证线程安全;
3. 使用重命名的方式进行滚动文件处理,如果失败会尝试一次不同分区的文件复制操作;
4. 删除过期日志有两个时机,一个是判断当前周期前 n 个周期文件,如果有则删除;
5. 对于设置了最大文件大小限制时,另外进行允许周期内的文件大小判定,超过大小后按修改时间最早删除;
6. 触发滚动时机后,进行异步删除,一般不影响业务;
第三方工具如: 经典版 cronolog, 时尚版 logrotate(麻烦)
cronolog 是一个很古老的日志滚动工具了(应该已经不维护了)。它可以接收应用的输出日志,然后按照规则进行日志存储,比如按照年月日时分秒来保存文件!
在网上其资料也已经不是很多了,很多人为了下载一个安装包也是绞尽脑汁啊!我也提供一个便捷安装包吧: 点此下载;
其 github 项目地址: https://github.com/fordmason/cronolog , 你完全可以自己去下载一个完全的包,自己安装!
不过我还是要说一下其他两个安装方式:
1. 直接使用 yum 源安装;(好像是要安装 epel 源) (推荐)
yum install cronolog -y
2. 使用上面下载的包,直接解压即可
tar -zxvf cronolog-bin.tar.gz -C /
3. 使用网上别人提供的源码安装
hehe...
说了这么多,还不是为了使用,如何与应用结合?
其实只需要在你原来应用启动的后面再加上如下命令就可以了!
$> | /usr/local/sbin/cronolog -S /var/logs/ai_ln.out /var/logs/ai.%Y-%m-%d-%H.out
完整的操作示例如下:
exec nohup java -jar /www/aproj\.jar 2>&1 | /usr/local/sbin/cronolog -S /var/logs/ai_ln.out /var/logs/ai.%Y-%m-%d-%H.out >> /dev/null &
如上命令是网上大部分人是这么写的,但是在某些情况下会有问题。比如我想远程启动这个服务的时候,就会一直拿不到结果!为啥?反正写成下面这个就完美了!即在 cronolog 之后,再加一个重定向输出 2>&1 。
exec nohup java -jar /www/aproj\.jar 2>&1 | /usr/local/sbin/cronolog -S /var/logs/ai_ln.out /var/logs/ai.%Y-%m-%d-%H.out >> /dev/null 2>&1 &
那么,这个工具和应用自己输出日志相比,有什么好处吗?它是怎么实现的呢?
好处前面已经说了,对代码无侵入,控制更灵活!
其实现原理为,接收一个标准的输入流,然后写入到相应文件即可!它不负责文件的删除,所以删除过期文件还得依赖另外的脚本!
其主体源码如下:
/* Loop, waiting for data on standard input */ for (;;) { /** * Read a buffer's worth of log file data, exiting on errors * or end of file. */ n_bytes_read = read(0, read_buf, sizeof read_buf); if (n_bytes_read == 0) { exit(3); } if (errno == EINTR) { continue; } else if (n_bytes_read < 0) { exit(4); } time_now = time(NULL) + time_offset; /** * If the current period has finished and there is a log file * open, close the log file */ if ((time_now >= next_period) && (log_fd >= 0)) { close(log_fd); log_fd = -1; } /** * If there is no log file open then open a new one. */ if (log_fd < 0) { log_fd = new_log_file(template, linkname, linktype, prevlinkname, periodicity, period_multiple, period_delay, filename, sizeof (filename), time_now, &next_period); } DEBUG(("%s (%d): wrote message; next period starts at %s (%d) in %d secs\n", timestamp(time_now), time_now, timestamp(next_period), next_period, next_period - time_now)); /** * Write out the log data to the current log file. */ if (write(log_fd, read_buf, n_bytes_read) != n_bytes_read) { perror(filename); exit(5); } }
大概操作就是:
1. cronolog 进程开启后,会一直死循环,除非遇到错误如应用关闭等;
2. 阻塞从标准输入读取信息,读取到后,再进行文件操作;
3. 每次读取内容后判断是否到达需要新滚动的周期,如果到了,就把原来的文件close掉,并重新创建一个用于写的文件;
4. 只管向打开的文件中写入缓冲内容即可;
5. 所有读入数据是基于管道操作的,简单实用;
看起来很简单啊!会不会有什么问题呢?应该不会吧,它可是经过时间考验的哦。越是简单的,往往越是可靠的!
看着上面代码,有同学肯定要说了,这么简单的代码谁不会啊,自己顺手就来一个shell搞定。 且不论你的shell写得是否可靠,但是你基于 shell, 别人是基于c的,恐怕不是一个量级的哦!
最后,还有个问题我们要处理下,那就是过期日志的清理问题?
这个简单的脚本是不会给你做了,或者说我没有发现它有这功能;所以,只能自己写脚本清理了!一行代码搞定!
# vim clean_log.sh find /var/logs/ai -mtime +8 -name "ai.*out" -exec rm -rf {} \; # 然后在 crontab 中加入执行时机即可,一般一天一次! 0 0 * * * sh clean_log.sh
搞定!
以上,就是一些日志滚动的实现及原理解析了!是不是有一种豁然开朗的感觉?哈哈。。
事情其实并没有想像中的难!