Examples with TaskAttemptID org.apache.hadoop.mapred.TaskAttemptID used on opensource projects

Example 16 with TaskAttemptID

use of org.apache.hadoop.mapred.TaskAttemptID in project incubator-systemml by apache.

the class MultipleOutputCommitter method commitTask.

@Override
public void commitTask(TaskAttemptContext context) throws IOException {
    JobConf conf = context.getJobConf();
    TaskAttemptID attemptId = context.getTaskAttemptID();
    // get the mapping between index to output filename
    outputs = MRJobConfiguration.getOutputs(conf);
    // get temp task output path (compatible with hadoop1 and hadoop2)
    Path taskOutPath = FileOutputFormat.getWorkOutputPath(conf);
    FileSystem fs = taskOutPath.getFileSystem(conf);
    if (!fs.exists(taskOutPath))
        throw new IOException("Task output path " + taskOutPath.toString() + "does not exist.");
    // move the task outputs to their final places
    context.getProgressible().progress();
    moveFinalTaskOutputs(context, fs, taskOutPath);
    // delete the temporary task-specific output directory
    if (!fs.delete(taskOutPath, true))
        LOG.debug("Failed to delete the temporary output directory of task: " + attemptId + " - " + taskOutPath);
}

Example 17 with TaskAttemptID

use of org.apache.hadoop.mapred.TaskAttemptID in project cdap by caskdata.

the class StreamInputFormatTest method testFormatStreamRecordReader.

@Test
public void testFormatStreamRecordReader() throws IOException, InterruptedException {
    File inputDir = tmpFolder.newFolder();
    File partition = new File(inputDir, "1.1000");
    partition.mkdirs();
    File eventFile = new File(partition, "bucket.1.0." + StreamFileType.EVENT.getSuffix());
    File indexFile = new File(partition, "bucket.1.0." + StreamFileType.INDEX.getSuffix());
    // write 1 event
    StreamDataFileWriter writer = new StreamDataFileWriter(Files.newOutputStreamSupplier(eventFile), Files.newOutputStreamSupplier(indexFile), 100L);
    StreamEvent streamEvent = new StreamEvent(ImmutableMap.of("header1", "value1", "header2", "value2"), Charsets.UTF_8.encode("hello world"), 1000);
    writer.append(streamEvent);
    writer.close();
    FormatSpecification formatSpec = new FormatSpecification(TextRecordFormat.class.getName(), Schema.recordOf("event", Schema.Field.of("body", Schema.of(Schema.Type.STRING))), Collections.<String, String>emptyMap());
    Configuration conf = new Configuration();
    AbstractStreamInputFormat.setStreamId(conf, DUMMY_ID);
    AbstractStreamInputFormat.setBodyFormatSpecification(conf, formatSpec);
    AbstractStreamInputFormat.setStreamPath(conf, inputDir.toURI());
    TaskAttemptContext context = new TaskAttemptContextImpl(conf, new TaskAttemptID());
    AbstractStreamInputFormat format = new AbstractStreamInputFormat() {

        @Override
        public AuthorizationEnforcer getAuthorizationEnforcer(TaskAttemptContext context) {
            return new NoOpAuthorizer();
        }

        @Override
        public AuthenticationContext getAuthenticationContext(TaskAttemptContext context) {
            return new AuthenticationTestContext();
        }
    };
    // read all splits and store the results in the list
    List<GenericStreamEventData<StructuredRecord>> recordsRead = Lists.newArrayList();
    List<InputSplit> inputSplits = format.getSplits(context);
    for (InputSplit split : inputSplits) {
        RecordReader<LongWritable, GenericStreamEventData<StructuredRecord>> recordReader = format.createRecordReader(split, context);
        recordReader.initialize(split, context);
        while (recordReader.nextKeyValue()) {
            recordsRead.add(recordReader.getCurrentValue());
        }
    }
    // should only have read 1 record
    Assert.assertEquals(1, recordsRead.size());
    GenericStreamEventData<StructuredRecord> eventData = recordsRead.get(0);
    Assert.assertEquals(streamEvent.getHeaders(), eventData.getHeaders());
    Assert.assertEquals("hello world", eventData.getBody().get("body"));
}