flume-hdfs 按照时间关闭并新开文件

sgdfh

在flume中 原有的sink的hdfs接口中只有根据写入多少秒存储关闭文档或者写入多少量存储关闭文档,饼不能根据实际需求来关闭文档,所以需要自己改写代码来满足我们的要求.例如 我们需要他在每天到了0点以后关闭上一日的文档,并在写一个新的文档出来,
首先我找到源码
在flume的源码中一个单独的项目flume-hdfs-sink是针对hdfs的可以只接在这个项目里进行修改,然后打包好在放到lib包中.
修改的类org.apache.flume.sink.hdfs.HDFSEventSink

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614

/** * Licensed to the Apache Software Foundation (ASF) under one * or more contributor license agreements.  See the NOTICE file * distributed with this work for additional information * regarding copyright ownership.  The ASF licenses this file * to you under the Apache License, Version 2.0 (the * "License"); you may not use this file except in compliance * with the License.  You may obtain a copy of the License at * *     http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ package org.apache.flume.sink.hdfs; import java.io.IOException; import java.util.ArrayList; import java.util.Calendar; import java.util.LinkedHashMap; import java.util.List; import java.util.Map; import java.util.TimeZone; import java.util.Map.Entry; import java.util.concurrent.ExecutorService; import java.util.concurrent.Executors; import java.util.concurrent.ScheduledExecutorService; import java.util.concurrent.TimeUnit; import com.google.common.annotations.VisibleForTesting; import org.apache.commons.lang.StringUtils; import org.apache.flume.Channel; import org.apache.flume.Clock; import org.apache.flume.Context; import org.apache.flume.Event; import org.apache.flume.EventDeliveryException; import org.apache.flume.SystemClock; import org.apache.flume.Transaction; import org.apache.flume.auth.FlumeAuthenticationUtil; import org.apache.flume.auth.FlumeAuthenticator; import org.apache.flume.auth.PrivilegedExecutor; import org.apache.flume.conf.Configurable; import org.apache.flume.formatter.output.BucketPath; import org.apache.flume.instrumentation.SinkCounter; import org.apache.flume.sink.AbstractSink; import org.apache.hadoop.conf.Configuration; import org.apache.hadoop.fs.FileSystem; import org.apache.hadoop.io.SequenceFile.CompressionType; import org.apache.hadoop.io.compress.CompressionCodec; import org.apache.hadoop.io.compress.CompressionCodecFactory; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import com.google.common.base.Preconditions; import com.google.common.collect.Lists; import com.google.common.util.concurrent.ThreadFactoryBuilder; public class HDFSEventSink extends AbstractSink implements Configurable {   public interface WriterCallback {     public void run(String filePath);   }   private static final Logger LOG = LoggerFactory       .getLogger(HDFSEventSink.class);   private static String DIRECTORY_DELIMITER = System.getProperty("file.separator");   private static final long defaultRollInterval = 30;   private static final long defaultRollSize = 1024;   private static final long defaultRollCount = 10;   private static final String defaultFileName = "FlumeData";   private static final String defaultSuffix = "";   private static final String defaultInUsePrefix = "";   private static final String defaultInUseSuffix = ".tmp";   private static final long defaultBatchSize = 100;   private static final String defaultFileType = HDFSWriterFactory.SequenceFileType;   private static final int defaultMaxOpenFiles = 5000;   // Time between close retries, in seconds   private static final long defaultRetryInterval = 180;   // Retry forever.   private static final int defaultTryCount = Integer.MAX_VALUE;   /**    * Default length of time we wait for blocking BucketWriter calls    * before timing out the operation. Intended to prevent server hangs.    */   private static final long defaultCallTimeout = 10000;   /**    * Default number of threads available for tasks    * such as append/open/close/flush with hdfs.    * These tasks are done in a separate thread in    * the case that they take too long. In which    * case we create a new file and move on.    */   private static final int defaultThreadPoolSize = 10;   private static final int defaultRollTimerPoolSize = 1;   private final HDFSWriterFactory writerFactory;   private WriterLinkedHashMap sfWriters;   private long rollInterval;   private long rollSize;   private long rollCount;   private long batchSize;   private int threadsPoolSize;   private int rollTimerPoolSize;   private CompressionCodec codeC;   private CompressionType compType;   private String fileType;   private String filePath;   private String fileName;   private String suffix;   private String inUsePrefix;   private String inUseSuffix;   private TimeZone timeZone;   private int maxOpenFiles;   private ExecutorService callTimeoutPool;   private ScheduledExecutorService timedRollerPool;   private boolean needRounding = false;   private int roundUnit = Calendar.SECOND;   private int roundValue = 1;   private boolean useLocalTime = false;   private long callTimeout;   private Context context;   private SinkCounter sinkCounter;   private volatile int idleTimeout;   private Clock clock;   private FileSystem mockFs;   private HDFSWriter mockWriter;   private final Object sfWritersLock = new Object();   private long retryInterval;   private int tryCount;   private PrivilegedExecutor privExecutor;   /**    * 根据文件前缀名生成文件滚动    */   private boolean rollName = false;   /*    * Extended Java LinkedHashMap for open file handle LRU queue.    * We want to clear the oldest file handle if there are too many open ones.    */   private static class WriterLinkedHashMap       extends LinkedHashMap<String, BucketWriter> {     private final int maxOpenFiles;     public WriterLinkedHashMap(int maxOpenFiles) {       super(16, 0.75f, true); // stock initial capacity/load, access ordering       this.maxOpenFiles = maxOpenFiles;     }     @Override     protected boolean removeEldestEntry(Entry<String, BucketWriter> eldest) {       if (size() > maxOpenFiles) {         // If we have more that max open files, then close the last one and         // return true         try {           eldest.getValue().close();         } catch (IOException e) {           LOG.warn(eldest.getKey().toString(), e);         } catch (InterruptedException e) {           LOG.warn(eldest.getKey().toString(), e);           Thread.currentThread().interrupt();         }         return true;       } else {         return false;       }     }   }   public HDFSEventSink() {     this(new HDFSWriterFactory());   }   public HDFSEventSink(HDFSWriterFactory writerFactory) {     this.writerFactory = writerFactory;   }   @VisibleForTesting   Map<String, BucketWriter> getSfWriters() {     return sfWriters;   }   // read configuration and setup thresholds   @Override   public void configure(Context context) {     this.context = context;     filePath = Preconditions.checkNotNull(         context.getString("hdfs.path"), "hdfs.path is required");     fileName = context.getString("hdfs.filePrefix", defaultFileName);     this.suffix = context.getString("hdfs.fileSuffix", defaultSuffix);     inUsePrefix = context.getString("hdfs.inUsePrefix", defaultInUsePrefix);     inUseSuffix = context.getString("hdfs.inUseSuffix", defaultInUseSuffix);     String tzName = context.getString("hdfs.timeZone");     timeZone = tzName == null ? null : TimeZone.getTimeZone(tzName);     rollInterval = context.getLong("hdfs.rollInterval", defaultRollInterval);     rollSize = context.getLong("hdfs.rollSize", defaultRollSize);     rollCount = context.getLong("hdfs.rollCount", defaultRollCount);     rollName = context.getBoolean("hdfs.rollName", false);     batchSize = context.getLong("hdfs.batchSize", defaultBatchSize);     idleTimeout = context.getInteger("hdfs.idleTimeout", 0);     String codecName = context.getString("hdfs.codeC");     fileType = context.getString("hdfs.fileType", defaultFileType);     maxOpenFiles = context.getInteger("hdfs.maxOpenFiles", defaultMaxOpenFiles);     callTimeout = context.getLong("hdfs.callTimeout", defaultCallTimeout);     threadsPoolSize = context.getInteger("hdfs.threadsPoolSize",         defaultThreadPoolSize);     rollTimerPoolSize = context.getInteger("hdfs.rollTimerPoolSize",         defaultRollTimerPoolSize);     String kerbConfPrincipal = context.getString("hdfs.kerberosPrincipal");     String kerbKeytab = context.getString("hdfs.kerberosKeytab");     String proxyUser = context.getString("hdfs.proxyUser");     tryCount = context.getInteger("hdfs.closeTries", defaultTryCount);     if(tryCount <= 0) {       LOG.warn("Retry count value : " + tryCount + " is not " +         "valid. The sink will try to close the file until the file " +         "is eventually closed.");       tryCount = defaultTryCount;     }     retryInterval = context.getLong("hdfs.retryInterval",       defaultRetryInterval);     if(retryInterval <= 0) {       LOG.warn("Retry Interval value: " + retryInterval + " is not " +         "valid. If the first close of a file fails, " +         "it may remain open and will not be renamed.");       tryCount = 1;     }     Preconditions.checkArgument(batchSize > 0,         "batchSize must be greater than 0");     if (codecName == null) {       codeC = null;       compType = CompressionType.NONE;     } else {       codeC = getCodec(codecName);       // TODO : set proper compression type       compType = CompressionType.BLOCK;     }     // Do not allow user to set fileType DataStream with codeC together     // To prevent output file with compress extension (like .snappy)     if(fileType.equalsIgnoreCase(HDFSWriterFactory.DataStreamType)         && codecName != null) {       throw new IllegalArgumentException("fileType: " + fileType +           " which does NOT support compressed output. Please don't set codeC" +           " or change the fileType if compressed output is desired.");     }     if(fileType.equalsIgnoreCase(HDFSWriterFactory.CompStreamType)) {       Preconditions.checkNotNull(codeC, "It's essential to set compress codec"           + " when fileType is: " + fileType);     }     // get the appropriate executor     this.privExecutor = FlumeAuthenticationUtil.getAuthenticator(             kerbConfPrincipal, kerbKeytab).proxyAs(proxyUser);     needRounding = context.getBoolean("hdfs.round", false);     if(needRounding) {       String unit = context.getString("hdfs.roundUnit", "second");       if (unit.equalsIgnoreCase("hour")) {         this.roundUnit = Calendar.HOUR_OF_DAY;       } else if (unit.equalsIgnoreCase("minute")) {         this.roundUnit = Calendar.MINUTE;       } else if (unit.equalsIgnoreCase("second")){         this.roundUnit = Calendar.SECOND;       } else {         LOG.warn("Rounding unit is not valid, please set one of" +             "minute, hour, or second. Rounding will be disabled");         needRounding = false;       }       this.roundValue = context.getInteger("hdfs.roundValue", 1);       if(roundUnit == Calendar.SECOND || roundUnit == Calendar.MINUTE){         Preconditions.checkArgument(roundValue > 0 && roundValue <= 60,             "Round value" +             "must be > 0 and <= 60");       } else if (roundUnit == Calendar.HOUR_OF_DAY){         Preconditions.checkArgument(roundValue > 0 && roundValue <= 24,             "Round value" +             "must be > 0 and <= 24");       }     }     this.useLocalTime = context.getBoolean("hdfs.useLocalTimeStamp", false);     if(useLocalTime) {       clock = new SystemClock();     }     if (sinkCounter == null) {       sinkCounter = new SinkCounter(getName());     }   }   private static boolean codecMatches(Class<? extends CompressionCodec> cls,       String codecName) {     String simpleName = cls.getSimpleName();     if (cls.getName().equals(codecName)         || simpleName.equalsIgnoreCase(codecName)) {       return true;     }     if (simpleName.endsWith("Codec")) {       String prefix = simpleName.substring(0,           simpleName.length() - "Codec".length());       if (prefix.equalsIgnoreCase(codecName)) {         return true;       }     }     return false;   }   @VisibleForTesting   static CompressionCodec getCodec(String codecName) {     Configuration conf = new Configuration();     List<Class<? extends CompressionCodec>> codecs = CompressionCodecFactory         .getCodecClasses(conf);     // Wish we could base this on DefaultCodec but appears not all codec's     // extend DefaultCodec(Lzo)     CompressionCodec codec = null;     ArrayList<String> codecStrs = new ArrayList<String>();     codecStrs.add("None");     for (Class<? extends CompressionCodec> cls : codecs) {       codecStrs.add(cls.getSimpleName());       if (codecMatches(cls, codecName)) {         try {           codec = cls.newInstance();         } catch (InstantiationException e) {           LOG.error("Unable to instantiate " + cls + " class");         } catch (IllegalAccessException e) {           LOG.error("Unable to access " + cls + " class");         }       }     }     if (codec == null) {       if (!codecName.equalsIgnoreCase("None")) {         throw new IllegalArgumentException("Unsupported compression codec "             + codecName + ".  Please choose from: " + codecStrs);       }     } else if (codec instanceof org.apache.hadoop.conf.Configurable) {       // Must check instanceof codec as BZip2Codec doesn't inherit Configurable       // Must set the configuration for Configurable objects that may or do use       // native libs       ((org.apache.hadoop.conf.Configurable) codec).setConf(conf);     }     return codec;   }   //TODO      private String oldName ="";   private void closeNowFile() {         for (Entry<String, BucketWriter> entry : sfWriters.entrySet()) {             String file = entry.getKey();             LOG.info("Closing {}", file);             try {                 entry.getValue().close();                 sfWriters.remove(file);             } catch (Exception ex) {                 LOG.warn("Exception while closing " + entry.getKey() + ". " + "Exception follows.",                         ex);                 if (ex instanceof InterruptedException) {                     Thread.currentThread().interrupt();                 }             }         }     }   /**    * Pull events out of channel and send it to HDFS. Take at most batchSize    * events per Transaction. Find the corresponding bucket for the event.    * Ensure the file is open. Serialize the data and write it to the file on    * HDFS. <br/>    * This method is not thread safe.    */   public Status process() throws EventDeliveryException {     if (rollName) {         String newName = BucketPath.escapeString(fileName, null,                   timeZone, needRounding, roundUnit, roundValue, useLocalTime);         if (StringUtils.isNotBlank(oldName)) {             if (!newName.equals(oldName)) {                 closeNowFile();             }         }         oldName =newName;     }            Channel channel = getChannel();     Transaction transaction = channel.getTransaction();     List<BucketWriter> writers = Lists.newArrayList();     transaction.begin();     try {       int txnEventCount = 0;       for (txnEventCount = 0; txnEventCount < batchSize; txnEventCount++) {         Event event = channel.take();         if (event == null) {           break;         }         // reconstruct the path name by substituting place holders         String realPath = BucketPath.escapeString(filePath, event.getHeaders(),             timeZone, needRounding, roundUnit, roundValue, useLocalTime);             String realName = BucketPath.escapeString(fileName, event.getHeaders(),               timeZone, needRounding, roundUnit, roundValue, useLocalTime);                      String lookupPath = realPath + DIRECTORY_DELIMITER + realName;         BucketWriter bucketWriter;         HDFSWriter hdfsWriter = null;         // Callback to remove the reference to the bucket writer from the         // sfWriters map so that all buffers used by the HDFS file         // handles are garbage collected.         WriterCallback closeCallback = new WriterCallback() {           @Override           public void run(String bucketPath) {             LOG.info("Writer callback called.");             synchronized (sfWritersLock) {               sfWriters.remove(bucketPath);             }           }         };         synchronized (sfWritersLock) {           bucketWriter = sfWriters.get(lookupPath);           // we haven't seen this file yet, so open it and cache the handle           if (bucketWriter == null) {             hdfsWriter = writerFactory.getWriter(fileType);             bucketWriter = initializeBucketWriter(realPath, realName,               lookupPath, hdfsWriter, closeCallback);             sfWriters.put(lookupPath, bucketWriter);           }         }         // track the buckets getting written in this transaction         if (!writers.contains(bucketWriter)) {           writers.add(bucketWriter);         }         //去掉信息的最后个回车         String infosStrings =new String(event.getBody());         if (infosStrings.length() == infosStrings.lastIndexOf("\n")+1) {             infosStrings =    infosStrings.substring(0,infosStrings.lastIndexOf("\n"));             event.setBody(infosStrings.getBytes());         }                   // Write the data to HDFS         try {           bucketWriter.append(event);         } catch (BucketClosedException ex) {           LOG.info("Bucket was closed while trying to append, " +             "reinitializing bucket and writing event.");           hdfsWriter = writerFactory.getWriter(fileType);           bucketWriter = initializeBucketWriter(realPath, realName,             lookupPath, hdfsWriter, closeCallback);           synchronized (sfWritersLock) {             sfWriters.put(lookupPath, bucketWriter);           }           bucketWriter.append(event);         }       }       if (txnEventCount == 0) {         sinkCounter.incrementBatchEmptyCount();       } else if (txnEventCount == batchSize) {         sinkCounter.incrementBatchCompleteCount();       } else {         sinkCounter.incrementBatchUnderflowCount();       }       // flush all pending buckets before committing the transaction       for (BucketWriter bucketWriter : writers) {         bucketWriter.flush();       }       transaction.commit();       if (txnEventCount < 1) {         return Status.BACKOFF;       } else {         sinkCounter.addToEventDrainSuccessCount(txnEventCount);         return Status.READY;       }     } catch (IOException eIO) {       transaction.rollback();       LOG.warn("HDFS IO error", eIO);       return Status.BACKOFF;     } catch (Throwable th) {       transaction.rollback();       LOG.error("process failed", th);       if (th instanceof Error) {         throw (Error) th;       } else {         throw new EventDeliveryException(th);       }     } finally {       transaction.close();     }   }   private BucketWriter initializeBucketWriter(String realPath,     String realName, String lookupPath, HDFSWriter hdfsWriter,     WriterCallback closeCallback) {     BucketWriter bucketWriter = new BucketWriter(rollInterval,       rollSize, rollCount,       batchSize, context, realPath, realName, inUsePrefix, inUseSuffix,       suffix, codeC, compType, hdfsWriter, timedRollerPool,       privExecutor, sinkCounter, idleTimeout, closeCallback,       lookupPath, callTimeout, callTimeoutPool, retryInterval,       tryCount);     if(mockFs != null) {       bucketWriter.setFileSystem(mockFs);       bucketWriter.setMockStream(mockWriter);     }     return bucketWriter;   }   @Override   public void stop() {     // do not constrain close() calls with a timeout     synchronized (sfWritersLock) {       for (Entry<String, BucketWriter> entry : sfWriters.entrySet()) {         LOG.info("Closing {}", entry.getKey());         try {           entry.getValue().close();         } catch (Exception ex) {           LOG.warn("Exception while closing " + entry.getKey() + ". " +                   "Exception follows.", ex);           if (ex instanceof InterruptedException) {             Thread.currentThread().interrupt();           }         }       }     }     // shut down all our thread pools     ExecutorService toShutdown[] = {callTimeoutPool, timedRollerPool};     for (ExecutorService execService : toShutdown) {       execService.shutdown();       try {         while (execService.isTerminated() == false) {           execService.awaitTermination(                   Math.max(defaultCallTimeout, callTimeout), TimeUnit.MILLISECONDS);         }       } catch (InterruptedException ex) {         LOG.warn("shutdown interrupted on " + execService, ex);       }     }     callTimeoutPool = null;     timedRollerPool = null;     synchronized (sfWritersLock) {       sfWriters.clear();       sfWriters = null;     }     sinkCounter.stop();     super.stop();   }   @Override   public void start() {     String timeoutName = "hdfs-" + getName() + "-call-runner-%d";     callTimeoutPool = Executors.newFixedThreadPool(threadsPoolSize,             new ThreadFactoryBuilder().setNameFormat(timeoutName).build());     String rollerName = "hdfs-" + getName() + "-roll-timer-%d";     timedRollerPool = Executors.newScheduledThreadPool(rollTimerPoolSize,             new ThreadFactoryBuilder().setNameFormat(rollerName).build());     this.sfWriters = new WriterLinkedHashMap(maxOpenFiles);     sinkCounter.start();     super.start();   }   @Override   public String toString() {     return "{ Sink type:" + getClass().getSimpleName() + ", name:" + getName() +             " }";   }   @VisibleForTesting   void setBucketClock(Clock clock) {     BucketPath.setClock(clock);   }   @VisibleForTesting   void setMockFs(FileSystem mockFs) {     this.mockFs = mockFs;   }   @VisibleForTesting   void setMockWriter(HDFSWriter writer) {     this.mockWriter = writer;   }   @VisibleForTesting   int getTryCount() {     return tryCount;   } }

在config中新增一个参数

1 2 3 4 5 6 7 8 9 10 11 12 13

home.sinks.r1.type = hdfs home.sinks.r1.channel = c1 home.sinks.r1.hdfs.path = hdfs://172.24.137.222/flume/stat/%y%m home.sinks.r1.hdfs.filePrefix = data-%y-%m-%d home.sinks.r1.hdfs.fileType = DataStream #开启使用动态日期格式 home.sinks.r1.hdfs.useLocalTimeStamp = true #0为关闭一条数更换文件 home.sinks.r1.hdfs.rollInterval = 0 home.sinks.r1.hdfs.rollCount = 0 home.sinks.r1.hdfs.rollSize = 0 #新增更换文件开关 home.sinks.r1.hdfs.rollName = true

前缀名中以动态日期的格式生成文件

在 process()  方法开始的时候添加 一个时间标记,对当前日期与文件创建日期判断,如日期有所改变,则改变之前的文件并新建新的文件,更改时间标记.

这样在hdfs中就会每天都生成一个文件