SpoolDirectorySource使用及源码分析
2015-11-07 16:58
387 查看
Spooling Directory Source简介
Spooling Directory Source可以获取硬盘上“spooling”目录的数据,这个Source将监视指定目录是否有新文件,如果有新文件的话,就解析这个新文件。事件的解析逻辑是可插拔的。在文件的内容所有的都读取到Channel之后,Spooling Directory Source会重名或者是删除该文件以表示文件已经读取完成。不像Exec Source,这个Source是可靠的,且不会丢失数据。即使Flume重启或者被Kill。但是需要注意如下两点:
1,如果文件在放入spooling目录之后还在写,那么Flume会打印错误日志,并且停止处理该文件。
2,如果文件之后重复使用,Flume将打印错误日志,并且停止处理。
为了避免以上问题,我们可以使用唯一的标识符来命令文件,例如:时间戳。
尽管这个Source是可靠的,但是如果下游发生故障,也会导致Event重复,这种情况就需要通过Flume的其他组件提供保障了。
属性名 | 默认 | 描述 |
---|---|---|
channels | – | |
type | – | 组件名:spooldir. |
spoolDir | – | 读取文件的目录。 |
fileSuffix | .COMPLETED | Spooling读取过的文件,添加的后缀。 |
deletePolicy | never | 完成后的文件是否删除。never:不删除或 immediate:立即删除 |
fileHeader | false | 是不把路径加入到Heander |
fileHeaderKey | file | 路径加入到Header的Key是什么 |
basenameHeader | false | 是不把文件名加入到Heander |
basenameHeaderKey | basename | 文件名加入到Header的Key是什么 |
ignorePattern | ^$ | 采用正则表达是去过滤一些文件。只有符合正则表达式的文件才会被使用。 |
trackerDir | .flumespool | 被处理文件的元数据的存储目录,如果不是绝对路径,就被会解析到spoolDir目录下。 |
consumeOrder | oldest | 消费spooling目录文件的规则,分别有:oldest,youngest和random。在oldest 和 youngest的情况下, 通过文件的最后修改时间来比较文件。如果最后修改时间相同,就根据字典的序列从小开始。在随机的情况 下,就随意读取文件。如果文件列表很长,采用oldest/youngest可能会很慢,因为用oldest/youngest要 扫描文件。但是如果采用random的话,就可能造成新的文件消耗的很快,老的文件一直都没有被消费。 |
maxBackoff | 4000 | 如果Channel已经满了,那么该Source连续尝试写入该Channel的最长时间(单位:毫秒)。 |
batchSize | 100 | 批量传输到Channel的粒度。 |
inputCharset | UTF-8 | 字符集 |
decodeErrorPolicy | FAIL | 在文件中有不可解析的字符时的解析策略。FAIL: 抛出一个异常,并且不能解析该文件。 REPLACE: 取代不可 解析的字符,通常用Unicode U+FFFD. IGNORE: 丢弃不可能解析字符序列。 |
deserializer | LINE | 自定序列化的方式,自定的话,必须实现EventDeserializer.Builder. |
deserializer.* | ||
bufferMaxLines | – | 已废弃。 |
bufferMaxLineLength | 5000 | (不推荐使用) 一行中最大的长度,可以使用deserializer.maxLineLength代替。 |
selector.type | replicating | replicating(复制) 或 multiplexing(复用) |
selector.* | 取决于selector.type的值 | |
interceptors | – | 空格分割的interceptor列表。 |
interceptors.* |
SpoolDirectorySource示例
读取文件写入到file_roll中a1.sources = source1 a1.sinks = sink1 a1.channels = channel1 #resources a1.sources.source1.type = spooldir a1.sources.source1.channels = channel1 a1.sources.source1.spoolDir = E:\\home\\spooling a1.sources.source1.fileHeader = true a1.sources.source1.fileHeaderKey = fishfile a1.sources.source1.basenameHeader = true a1.sources.source1.basenameHeaderKey = fishbasename a1.sinks.sink1.type = file_roll a1.sinks.sink1.sink.directory = E:\\home\\file_roll a1.sinks.sink1.sink.rollInterval = 300 a1.sinks.sink1.sink.serializer = TEXT a1.sinks.sink1.sink.batchSize = 100 a1.channels.channel1.type = memory a1.channels.channel1.capacity = 1000 a1.channels.channel1.transactionCapacity = 100 a1.sources.source1.channels = channel1 a1.sinks.sink1.channel = channel1
SpoolDirectorySource源码分析
一,调用configure(Context context)方法初始化:@Override public synchronized void configure(Context context) { //spool目录 spoolDirectory = context.getString(SPOOL_DIRECTORY); Preconditions.checkState(spoolDirectory != null, "Configuration must specify a spooling directory"); //完成后的文件后缀 completedSuffix = context.getString(SPOOLED_FILE_SUFFIX, DEFAULT_SPOOLED_FILE_SUFFIX); //删除策略,never:不删除 或 immediate:立即删除 deletePolicy = context.getString(DELETE_POLICY, DEFAULT_DELETE_POLICY); //以下四个参数是是否在header中加入文件名和文件路径。 fileHeader = context.getBoolean(FILENAME_HEADER, DEFAULT_FILE_HEADER); fileHeaderKey = context.getString(FILENAME_HEADER_KEY, DEFAULT_FILENAME_HEADER_KEY); basenameHeader = context.getBoolean(BASENAME_HEADER, DEFAULT_BASENAME_HEADER); basenameHeaderKey = context.getString(BASENAME_HEADER_KEY, DEFAULT_BASENAME_HEADER_KEY); //批量处理的数量 batchSize = context.getInteger(BATCH_SIZE, DEFAULT_BATCH_SIZE); //字符集 inputCharset = context.getString(INPUT_CHARSET, DEFAULT_INPUT_CHARSET); //在文件中有不可解析的字符时的解析策略 decodeErrorPolicy = DecodeErrorPolicy.valueOf( context.getString(DECODE_ERROR_POLICY, DEFAULT_DECODE_ERROR_POLICY) .toUpperCase(Locale.ENGLISH)); //过滤文件的正则表达式 ignorePattern = context.getString(IGNORE_PAT, DEFAULT_IGNORE_PAT); //被处理文件的元数据的存储目录 trackerDirPath = context.getString(TRACKER_DIR, DEFAULT_TRACKER_DIR); //序列化 deserializerType = context.getString(DESERIALIZER, DEFAULT_DESERIALIZER); deserializerContext = new Context(context.getSubProperties(DESERIALIZER + ".")); //消费spooling目录文件的规则 consumeOrder = ConsumeOrder.valueOf(context.getString(CONSUME_ORDER, DEFAULT_CONSUME_ORDER.toString()).toUpperCase(Locale.ENGLISH)); // "Hack" to support backwards compatibility with previous generation of // spooling directory source, which did not support deserializers Integer bufferMaxLineLength = context.getInteger(BUFFER_MAX_LINE_LENGTH); if (bufferMaxLineLength != null && deserializerType != null && deserializerType.equalsIgnoreCase(DEFAULT_DESERIALIZER)) { deserializerContext.put(LineDeserializer.MAXLINE_KEY, bufferMaxLineLength.toString()); } maxBackoff = context.getInteger(MAX_BACKOFF, DEFAULT_MAX_BACKOFF); if (sourceCounter == null) { sourceCounter = new SourceCounter(getName()); } }
start方法:
@Override public synchronized void start() { logger.info("SpoolDirectorySource source starting with directory: {}", spoolDirectory); executor = Executors.newSingleThreadScheduledExecutor(); File directory = new File(spoolDirectory); //构建ReliableSpoolingFileEventReader对象 try { reader = new ReliableSpoolingFileEventReader.Builder() .spoolDirectory(directory) .completedSuffix(completedSuffix) .ignorePattern(ignorePattern) .trackerDirPath(trackerDirPath) .annotateFileName(fileHeader) .fileNameHeader(fileHeaderKey) .annotateBaseName(basenameHeader) .baseNameHeader(basenameHeaderKey) .deserializerType(deserializerType) .deserializerContext(deserializerContext) .deletePolicy(deletePolicy) .inputCharset(inputCharset) .decodeErrorPolicy(decodeErrorPolicy) .consumeOrder(consumeOrder) .build(); } catch (IOException ioe) { throw new FlumeException("Error instantiating spooling event parser", ioe); } //构建SpoolDirectoryRunnable线程。 Runnable runner = new SpoolDirectoryRunnable(reader, sourceCounter); //每隔POLL_DELAY_MS(500ms)执行以下SpoolDirectoryRunnable线程。 executor.scheduleWithFixedDelay( runner, 0, POLL_DELAY_MS, TimeUnit.MILLISECONDS); super.start(); logger.debug("SpoolDirectorySource source started"); sourceCounter.start(); }构建ReliableSpoolingFileEventReader类,构造方法功能:
1,spooling目录是否存在,是否是目录
2,通过创建零时文件测试spooling目录的权限。
3,创建trackerDir目录和.flumespool-main.meta文件
SpoolDirectoryRunnable线程,主要用于发送和读取Event:
private class SpoolDirectoryRunnable implements Runnable { private ReliableSpoolingFileEventReader reader; private SourceCounter sourceCounter; public SpoolDirectoryRunnable(ReliableSpoolingFileEventReader reader, SourceCounter sourceCounter) { this.reader = reader; this.sourceCounter = sourceCounter; } @Override public void run() { int backoffInterval = 250; try { while (!Thread.interrupted()) { //ReliableSpoolingFileEventReader读取batchSize大小的Event List<Event> events = reader.readEvents(batchSize); if (events.isEmpty()) { break; } //统计 sourceCounter.addToEventReceivedCount(events.size()); sourceCounter.incrementAppendBatchReceivedCount(); try { //将Event数组发送到Channel getChannelProcessor().processEventBatch(events); //commit会记录最后一次读取的行数,以便下次知道从哪里开始读 reader.commit(); } catch (ChannelException ex) { //ChannelProcessor批量提交Event出错,会抛出ChannelException异常,此时reader.commit是没有执行的 //所以在接下来的continue后,继续通过reader读取文件的话,还是从原来的位置读取,以保证数据不会丢失。 logger.warn("The channel is full, and cannot write data now. The " + "source will try again after " + String.valueOf(backoffInterval) + " milliseconds"); hitChannelException = true; if (backoff) { TimeUnit.MILLISECONDS.sleep(backoffInterval); backoffInterval = backoffInterval << 1; backoffInterval = backoffInterval >= maxBackoff ? maxBackoff : backoffInterval; } continue; } backoffInterval = 250; sourceCounter.addToEventAcceptedCount(events.size()); sourceCounter.incrementAppendBatchAcceptedCount(); } } catch (Throwable t) { logger.error("FATAL: " + SpoolDirectorySource.this.toString() + ": " + "Uncaught exception in SpoolDirectorySource thread. " + "Restart or reconfigure Flume to continue processing.", t); hasFatalError = true; Throwables.propagate(t); } } }ReliableSpoolingFileEventReader读取Event:
public List<Event> readEvents(int numEvents) throws IOException { //committed初始化为true if (!committed) { if (!currentFile.isPresent()) { throw new IllegalStateException("File should not roll when " + "commit is outstanding."); } logger.info("Last read was never committed - resetting mark position."); //正常情况下,会在SpoolDirectorySource类中记录读取的字节数之后,将commited设置为true //没有设置为true,可能是因为发送到Channel异常了,调用下面reset方法可以保证数据不丢失。 currentFile.get().getDeserializer().reset(); } else { // Check if new files have arrived since last call if (!currentFile.isPresent()) { //读取文件,读取文件过程中使用FileFilter过滤掉completedSuffix后缀的文件,然后根据消费文件的规则(consumeOrder)去消费文件。 currentFile = getNextFile(); } // Return empty list if no new files if (!currentFile.isPresent()) { return Collections.emptyList(); } } EventDeserializer des = currentFile.get().getDeserializer(); //根据序列化类读取Event List<Event> events = des.readEvents(numEvents); /* It's possible that the last read took us just up to a file boundary. * If so, try to roll to the next file, if there is one. * Loop until events is not empty or there is no next file in case of 0 byte files */ while (events.isEmpty()) { logger.info("Last read took us just up to a file boundary. Rolling to the next file, if there is one."); retireCurrentFile(); currentFile = getNextFile(); if (!currentFile.isPresent()) { return Collections.emptyList(); } events = currentFile.get().getDeserializer().readEvents(numEvents); } //添加<span style="font-family:Microsoft Yahei;">文件路径到Header</span> if (annotateFileName) { String filename = currentFile.get().getFile().getAbsolutePath(); for (Event event : events) { event.getHeaders().put(fileNameHeader, filename); } } //添加文件名到Header if (annotateBaseName) { String basename = currentFile.get().getFile().getName(); for (Event event : events) { event.getHeaders().put(baseNameHeader, basename); } } committed = false; lastFileRead = currentFile; return events; }消费文件的规则,以OLDEST为例:
for (File candidateFile: candidateFiles) { //已选择文件的最后修改时间,减去文件列表取的文件最后修改时间 long compare = selectedFile.lastModified() - candidateFile.lastModified(); if (compare == 0) { // ts is same pick smallest lexicographically. //时间一样就根据字典序列排序 selectedFile = smallerLexicographical(selectedFile, candidateFile); } else if (compare > 0) { // candidate is older (cand-ts < selec-ts). selectedFile = candidateFile; } }
在ReliableSpoolingFileEventReader类读取Events之后,调用commit方法提交,这里有个问题,batchSize为200,我们每次也就读取200行,那么SpoolingDirectorySource是如何标记我们读到文件的那个位置的呢?
其实SpoolingDirectorySource在commit时,会调用EventDeserializer的mark方法标记此时读取在文件的什么位置。源码如下:
public synchronized void storePosition(long position) throws IOException { metaCache.setOffset(position); writer.append(metaCache); writer.sync(); writer.flush(); }上面的方法会把文件的读取到什么位置记录到.flumespool\.flumespool-main.meta(默认情况下)文件中。
public void commit() throws IOException { if (!committed && currentFile.isPresent()) { currentFile.get().getDeserializer().mark(); //mark成功后,将committed设置为true committed = true; } }
相关文章推荐
- Flume环境部署和配置详解及案例大全
- Play! Akka Flume实现的完整数据收集
- flume自定义Interceptor
- #Note# Analyzing Twitter Data with Apache Hadoo...
- flume、kafka、storm常用命令
- 开源日志系统比较
- Flume向HDFS写数据实例
- flume+log4j整合到web项目
- 详细图解 Flume介绍、安装配置-1
- flume部署
- flume实时抓取log数据,并传到kafka中
- flume NG 中文 Welcome to Apache Flume 第一页 醉了
- flume 高可用性 高可靠性 agent source
- flume介绍及扩展开发心得
- Flume
- Flume-1.6.0修改
- Log4j2的flume appender配置
- flume-两台机器上agent的串联运行
- log4j+flume+kafka管理日志,查询日志
- 实现flume tailfsource解决丢数等问题