List of usage examples for org.apache.hadoop.mapreduce RecordWriter close
public abstract void close(TaskAttemptContext context) throws IOException, InterruptedException;
RecordWriter to future operations. From source file:co.cask.cdap.internal.app.runtime.batch.dataset.output.MultipleOutputs.java
License:Apache License
/** * Closes a collection of RecordWriters, suppressing any exceptions until close is called on each of them. * * @param recordWriters The Collection of RecordWriters to close * @param context The context to pass during close of each RecordWriter *//*from w w w .j a v a2s. c o m*/ public static void closeRecordWriters(Iterable<RecordWriter<?, ?>> recordWriters, TaskAttemptContext context) { RuntimeException ex = null; for (RecordWriter writer : recordWriters) { try { writer.close(context); } catch (IOException | InterruptedException e) { if (ex == null) { ex = new RuntimeException(e); } else { ex.addSuppressed(e); } } } if (ex != null) { throw ex; } }
From source file:com.asakusafw.runtime.mapreduce.simple.SimpleJobRunner.java
License:Apache License
@SuppressWarnings({ "rawtypes", "unchecked" })
private void runMap(Job job, KeyValueSorter<?, ?> sorter)
throws IOException, InterruptedException, ClassNotFoundException {
Configuration conf = job.getConfiguration();
InputFormat<?, ?> input = ReflectionUtils.newInstance(job.getInputFormatClass(), conf);
List<InputSplit> splits = input.getSplits(job);
int serial = 1;
for (InputSplit split : splits) {
TaskAttemptID id = newTaskAttemptId(newMapTaskId(job.getJobID(), serial++), 0);
Mapper<?, ?, ?, ?> mapper = ReflectionUtils.newInstance(job.getMapperClass(), conf);
if (LOG.isDebugEnabled()) {
LOG.debug(MessageFormat.format("starting mapper: {0}@{1} ({2}bytes)", //$NON-NLS-1$
mapper.getClass().getName(), id, split.getLength()));
}/*from ww w .j a va 2 s. co m*/
TaskAttemptContext context = newTaskAttemptContext(conf, id);
// we always obtain a new OutputFormat object / OutputFormat.getOutputCommiter() may be cached
OutputFormat<?, ?> output = ReflectionUtils.newInstance(job.getOutputFormatClass(), conf);
OutputCommitter committer = output.getOutputCommitter(context);
committer.setupTask(context);
boolean succeed = false;
try (RecordReader<?, ?> reader = input.createRecordReader(split, newTaskAttemptContext(conf, id))) {
RecordWriter<?, ?> writer;
if (sorter != null) {
writer = new ShuffleWriter(sorter);
} else {
writer = output.getRecordWriter(newTaskAttemptContext(conf, id));
}
try {
Mapper.Context c = newMapperContext(conf, id, reader, writer, committer, split);
reader.initialize(split, c);
mapper.run(c);
} finally {
writer.close(newTaskAttemptContext(conf, id));
}
doCommitTask(context, committer);
succeed = true;
} finally {
if (succeed == false) {
doAbortTask(context, committer);
}
}
}
}
From source file:com.asakusafw.runtime.mapreduce.simple.SimpleJobRunner.java
License:Apache License
@SuppressWarnings({ "unchecked", "rawtypes" })
private void runReduce(Job job, KeyValueSorter<?, ?> sorter)
throws ClassNotFoundException, IOException, InterruptedException {
Configuration conf = job.getConfiguration();
OutputFormat<?, ?> output = ReflectionUtils.newInstance(job.getOutputFormatClass(), conf);
TaskAttemptID id = newTaskAttemptId(newReduceTaskId(job.getJobID(), 1), 0);
Reducer<?, ?, ?, ?> reducer = ReflectionUtils.newInstance(job.getReducerClass(), conf);
if (LOG.isDebugEnabled()) {
LOG.debug(MessageFormat.format("starting reducer: {0}@{1} ({2}records, {3}bytes)", //$NON-NLS-1$
reducer.getClass().getName(), id, sorter.getRecordCount(), sorter.getSizeInBytes()));
}/*from www .j av a2s . com*/
TaskAttemptContext context = newTaskAttemptContext(conf, id);
OutputCommitter committer = output.getOutputCommitter(context);
committer.setupTask(context);
boolean succeed = false;
try {
ShuffleReader reader = new ShuffleReader(sorter, new Progress());
try {
RecordWriter<?, ?> writer = output.getRecordWriter(newTaskAttemptContext(conf, id));
try {
Reducer.Context c = newReducerContext(conf, id, reader, sorter.getKeyClass(),
sorter.getValueClass(), writer, committer, (RawComparator) job.getGroupingComparator());
reducer.run(c);
} finally {
writer.close(newTaskAttemptContext(conf, id));
}
} finally {
try {
reader.close();
} catch (IOException e) {
LOG.warn(MessageFormat.format("error occurred while reducer mapper input: {0} ({1})", id,
job.getJobName()), e);
}
}
doCommitTask(context, committer);
succeed = true;
} finally {
if (succeed == false) {
doAbortTask(context, committer);
}
}
}
From source file:com.facebook.hiveio.output.OutputCmd.java
License:Apache License
/** * Write output// w w w.ja v a 2 s . c om * * @param context Context * @throws Exception */ public void write(Context context) throws Exception { PerThread threadLocal = context.perThread.get(); HiveApiOutputCommitter outputCommitter = context.outputFormat.getOutputCommitter(threadLocal.taskContext()); outputCommitter.setupTask(threadLocal.taskContext()); RecordWriter<WritableComparable, HiveWritableRecord> recordWriter = context.outputFormat .getRecordWriter(threadLocal.taskContext()); HiveWritableRecord record = HiveRecordFactory.newWritableRecord(context.schema); // TODO: allow type promotions: see https://github.com/facebook/hive-io-experimental/issues/15 record.set(0, 11L); record.set(1, 22.22); record.set(2, true); record.set(3, "foo"); recordWriter.write(NullWritable.get(), record); record.set(0, 33L); record.set(1, 44.44); record.set(2, false); record.set(3, "bar"); recordWriter.write(NullWritable.get(), record); recordWriter.close(threadLocal.taskContext()); if (outputCommitter.needsTaskCommit(threadLocal.taskContext())) { outputCommitter.commitTask(threadLocal.taskContext()); } }
From source file:com.github.dryangkun.hbase.tidx.hive.HiveHFileOutputFormat.java
License:Apache License
@Override public RecordWriter getHiveRecordWriter(final JobConf jc, final Path finalOutPath, Class<? extends Writable> valueClass, boolean isCompressed, Properties tableProperties, final Progressable progressable) throws IOException { // Read configuration for the target path, first from jobconf, then from table properties String hfilePath = getFamilyPath(jc, tableProperties); if (hfilePath == null) { throw new RuntimeException("Please set " + HFILE_FAMILY_PATH + " to target location for HFiles"); }//from w w w . jav a 2 s . c o m // Target path's last component is also the column family name. final Path columnFamilyPath = new Path(hfilePath); final String columnFamilyName = columnFamilyPath.getName(); final byte[] columnFamilyNameBytes = Bytes.toBytes(columnFamilyName); final Job job = new Job(jc); setCompressOutput(job, isCompressed); setOutputPath(job, finalOutPath); // Create the HFile writer final org.apache.hadoop.mapreduce.TaskAttemptContext tac = ShimLoader.getHadoopShims() .newTaskAttemptContext(job.getConfiguration(), progressable); final Path outputdir = FileOutputFormat.getOutputPath(tac); final org.apache.hadoop.mapreduce.RecordWriter<ImmutableBytesWritable, KeyValue> fileWriter = getFileWriter( tac); // Individual columns are going to be pivoted to HBase cells, // and for each row, they need to be written out in order // of column name, so sort the column names now, creating a // mapping to their column position. However, the first // column is interpreted as the row key. String columnList = tableProperties.getProperty("columns"); String[] columnArray = columnList.split(","); final SortedMap<byte[], Integer> columnMap = new TreeMap<byte[], Integer>(Bytes.BYTES_COMPARATOR); int i = 0; for (String columnName : columnArray) { if (i != 0) { columnMap.put(Bytes.toBytes(columnName), i); } ++i; } return new RecordWriter() { @Override public void close(boolean abort) throws IOException { try { fileWriter.close(null); if (abort) { return; } // Move the hfiles file(s) from the task output directory to the // location specified by the user. FileSystem fs = outputdir.getFileSystem(jc); fs.mkdirs(columnFamilyPath); Path srcDir = outputdir; for (;;) { FileStatus[] files = fs.listStatus(srcDir, FileUtils.STAGING_DIR_PATH_FILTER); if ((files == null) || (files.length == 0)) { throw new IOException("No family directories found in " + srcDir); } if (files.length != 1) { throw new IOException("Multiple family directories found in " + srcDir); } srcDir = files[0].getPath(); if (srcDir.getName().equals(columnFamilyName)) { break; } } for (FileStatus regionFile : fs.listStatus(srcDir, FileUtils.STAGING_DIR_PATH_FILTER)) { fs.rename(regionFile.getPath(), new Path(columnFamilyPath, regionFile.getPath().getName())); } // Hive actually wants a file as task output (not a directory), so // replace the empty directory with an empty file to keep it happy. fs.delete(outputdir, true); fs.createNewFile(outputdir); } catch (InterruptedException ex) { throw new IOException(ex); } } private void writeText(Text text) throws IOException { // Decompose the incoming text row into fields. String s = text.toString(); String[] fields = s.split("\u0001"); assert (fields.length <= (columnMap.size() + 1)); // First field is the row key. byte[] rowKeyBytes = Bytes.toBytes(fields[0]); // Remaining fields are cells addressed by column name within row. for (Map.Entry<byte[], Integer> entry : columnMap.entrySet()) { byte[] columnNameBytes = entry.getKey(); int iColumn = entry.getValue(); String val; if (iColumn >= fields.length) { // trailing blank field val = ""; } else { val = fields[iColumn]; if ("\\N".equals(val)) { // omit nulls continue; } } byte[] valBytes = Bytes.toBytes(val); KeyValue kv = new KeyValue(rowKeyBytes, columnFamilyNameBytes, columnNameBytes, valBytes); try { fileWriter.write(null, kv); } catch (IOException e) { LOG.error("Failed while writing row: " + s); throw e; } catch (InterruptedException ex) { throw new IOException(ex); } } } private void writePut(PutWritable put) throws IOException { ImmutableBytesWritable row = new ImmutableBytesWritable(put.getPut().getRow()); SortedMap<byte[], List<Cell>> cells = put.getPut().getFamilyCellMap(); for (Map.Entry<byte[], List<Cell>> entry : cells.entrySet()) { Collections.sort(entry.getValue(), new CellComparator()); for (Cell c : entry.getValue()) { try { fileWriter.write(row, KeyValueUtil.copyToNewKeyValue(c)); } catch (InterruptedException e) { throw (InterruptedIOException) new InterruptedIOException().initCause(e); } } } } @Override public void write(Writable w) throws IOException { if (w instanceof Text) { writeText((Text) w); } else if (w instanceof PutWritable) { writePut((PutWritable) w); } else { throw new IOException("Unexpected writable " + w); } } }; }
From source file:com.hadoop.mapreduce.TestLzoTextInputFormat.java
License:Open Source License
/** * Creates an lzo file with random data. * /*from w w w . jav a 2 s. c o m*/ * @param outputDir Output directory. * @param fs File system we're using. * @param attemptContext Task attempt context, contains task id etc. * @throws IOException * @throws InterruptedException */ private byte[] createTestInput(Path outputDir, FileSystem fs, TaskAttemptContext attemptContext, int charsToOutput) throws IOException, InterruptedException { TextOutputFormat<Text, Text> output = new TextOutputFormat<Text, Text>(); RecordWriter<Text, Text> rw = null; md5.reset(); try { rw = output.getRecordWriter(attemptContext); char[] chars = "abcdefghijklmnopqrstuvwxyz\u00E5\u00E4\u00F6".toCharArray(); Random r = new Random(System.currentTimeMillis()); Text key = new Text(); Text value = new Text(); int charsMax = chars.length - 1; for (int i = 0; i < charsToOutput;) { i += fillText(chars, r, charsMax, key); i += fillText(chars, r, charsMax, value); rw.write(key, value); md5.update(key.getBytes(), 0, key.getLength()); // text output format writes tab between the key and value md5.update("\t".getBytes("UTF-8")); md5.update(value.getBytes(), 0, value.getLength()); } } finally { if (rw != null) { rw.close(attemptContext); OutputCommitter committer = output.getOutputCommitter(attemptContext); committer.commitTask(attemptContext); committer.cleanupJob(attemptContext); } } byte[] result = md5.digest(); md5.reset(); return result; }
From source file:com.linkedin.pinot.hadoop.io.PinotOutputFormatTest.java
License:Apache License
private Map<Integer, Emp> addTestData() throws IOException, InterruptedException { int days = 2000; int sal = 20; RecordWriter<Object, Emp> writer = outputFormat.getRecordWriter(fakeTaskAttemptContext); Map<Integer, Emp> inputMap = new HashMap<>(); for (int i = 0; i < 10; i++) { String name = "name " + i; Emp e = new Emp(i, name, days + i, sal + i); writer.write(null, e);//w ww . ja v a 2 s .c om inputMap.put(i, e); } writer.close(fakeTaskAttemptContext); return inputMap; }
From source file:com.linkedin.whiteelephant.mapreduce.MyAvroMultipleOutputs.java
License:Apache License
/** * Closes all the opened outputs./* ww w .j av a2 s . c o m*/ * * This should be called from cleanup method of map/reduce task. * If overridden subclasses must invoke <code>super.close()</code> at the * end of their <code>close()</code> * */ @SuppressWarnings("unchecked") public void close() throws IOException, InterruptedException { for (RecordWriter writer : recordWriters.values()) { writer.close(context); } }
From source file:com.marklogic.contentpump.LocalJobRunner.java
License:Apache License
/** * Run the job. Get the input splits, create map tasks and submit it to * the thread pool if there is one; otherwise, runs the the task one by * one.// w w w .j a va 2 s.co m * * @param <INKEY> * @param <INVALUE> * @param <OUTKEY> * @param <OUTVALUE> * @throws Exception */ @SuppressWarnings("unchecked") public <INKEY, INVALUE, OUTKEY, OUTVALUE, T extends org.apache.hadoop.mapreduce.InputSplit> void run() throws Exception { Configuration conf = job.getConfiguration(); InputFormat<INKEY, INVALUE> inputFormat = (InputFormat<INKEY, INVALUE>) ReflectionUtils .newInstance(job.getInputFormatClass(), conf); List<InputSplit> splits = inputFormat.getSplits(job); T[] array = (T[]) splits.toArray(new org.apache.hadoop.mapreduce.InputSplit[splits.size()]); // sort the splits into order based on size, so that the biggest // goes first Arrays.sort(array, new SplitLengthComparator()); OutputFormat<OUTKEY, OUTVALUE> outputFormat = (OutputFormat<OUTKEY, OUTVALUE>) ReflectionUtils .newInstance(job.getOutputFormatClass(), conf); Class<? extends Mapper<?, ?, ?, ?>> mapperClass = job.getMapperClass(); Mapper<INKEY, INVALUE, OUTKEY, OUTVALUE> mapper = (Mapper<INKEY, INVALUE, OUTKEY, OUTVALUE>) ReflectionUtils .newInstance(mapperClass, conf); try { outputFormat.checkOutputSpecs(job); } catch (Exception ex) { if (LOG.isDebugEnabled()) { LOG.debug("Error checking output specification: ", ex); } else { LOG.error("Error checking output specification: "); LOG.error(ex.getMessage()); } return; } conf = job.getConfiguration(); progress = new AtomicInteger[splits.size()]; for (int i = 0; i < splits.size(); i++) { progress[i] = new AtomicInteger(); } Monitor monitor = new Monitor(); monitor.start(); reporter = new ContentPumpReporter(); List<Future<Object>> taskList = new ArrayList<Future<Object>>(); for (int i = 0; i < array.length; i++) { InputSplit split = array[i]; if (pool != null) { LocalMapTask<INKEY, INVALUE, OUTKEY, OUTVALUE> task = new LocalMapTask<INKEY, INVALUE, OUTKEY, OUTVALUE>( inputFormat, outputFormat, conf, i, split, reporter, progress[i]); availableThreads = assignThreads(i, array.length); Class<? extends Mapper<?, ?, ?, ?>> runtimeMapperClass = job.getMapperClass(); if (availableThreads > 1 && availableThreads != threadsPerSplit) { // possible runtime adjustment if (runtimeMapperClass != (Class) MultithreadedMapper.class) { runtimeMapperClass = (Class<? extends Mapper<INKEY, INVALUE, OUTKEY, OUTVALUE>>) cmd .getRuntimeMapperClass(job, mapperClass, threadsPerSplit, availableThreads); } if (runtimeMapperClass != mapperClass) { task.setMapperClass(runtimeMapperClass); } if (runtimeMapperClass == (Class) MultithreadedMapper.class) { task.setThreadCount(availableThreads); if (LOG.isDebugEnabled()) { LOG.debug("Thread Count for Split#" + i + " : " + availableThreads); } } } if (runtimeMapperClass == (Class) MultithreadedMapper.class) { synchronized (pool) { taskList.add(pool.submit(task)); pool.wait(); } } else { pool.submit(task); } } else { // single-threaded JobID jid = new JobID(); TaskID taskId = new TaskID(jid.getJtIdentifier(), jid.getId(), TaskType.MAP, i); TaskAttemptID taskAttemptId = new TaskAttemptID(taskId, 0); TaskAttemptContext context = ReflectionUtil.createTaskAttemptContext(conf, taskAttemptId); RecordReader<INKEY, INVALUE> reader = inputFormat.createRecordReader(split, context); RecordWriter<OUTKEY, OUTVALUE> writer = outputFormat.getRecordWriter(context); OutputCommitter committer = outputFormat.getOutputCommitter(context); TrackingRecordReader trackingReader = new TrackingRecordReader(reader, progress[i]); Mapper.Context mapperContext = ReflectionUtil.createMapperContext(mapper, conf, taskAttemptId, trackingReader, writer, committer, reporter, split); trackingReader.initialize(split, mapperContext); // no thread pool (only 1 thread specified) Class<? extends Mapper<?, ?, ?, ?>> mapClass = job.getMapperClass(); mapperContext.getConfiguration().setClass(CONF_MAPREDUCE_JOB_MAP_CLASS, mapClass, Mapper.class); mapper = (Mapper<INKEY, INVALUE, OUTKEY, OUTVALUE>) ReflectionUtils.newInstance(mapClass, mapperContext.getConfiguration()); mapper.run(mapperContext); trackingReader.close(); writer.close(mapperContext); committer.commitTask(context); } } // wait till all tasks are done if (pool != null) { for (Future<Object> f : taskList) { f.get(); } pool.shutdown(); while (!pool.awaitTermination(1, TimeUnit.DAYS)) ; jobComplete.set(true); } monitor.interrupt(); monitor.join(1000); // report counters Iterator<CounterGroup> groupIt = reporter.counters.iterator(); while (groupIt.hasNext()) { CounterGroup group = groupIt.next(); LOG.info(group.getDisplayName() + ": "); Iterator<Counter> counterIt = group.iterator(); while (counterIt.hasNext()) { Counter counter = counterIt.next(); LOG.info(counter.getDisplayName() + ": " + counter.getValue()); } } LOG.info("Total execution time: " + (System.currentTimeMillis() - startTime) / 1000 + " sec"); }
From source file:com.metamx.milano.hadoop.MilanoProtoFileOutputFormatTests.java
License:Apache License
@Test public void testBuildEmptyProtoFile() throws Exception { MilanoProtoFileOutputFormat outputFormat = new MilanoProtoFileOutputFormat(); @SuppressWarnings("unchecked") RecordWriter<String, Message> writer = outputFormat.getRecordWriter(protoTestObjects.getContext()); writer.close(protoTestObjects.getContext()); }