Examples with TaskAttemptContextImpl org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl used on opensource projects

Example 36 with TaskAttemptContextImpl

use of org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl in project hadoop by apache.

the class TestFileOutputCommitter method testFailAbortInternal.

private void testFailAbortInternal(int version) throws IOException, InterruptedException {
    Job job = Job.getInstance();
    Configuration conf = job.getConfiguration();
    conf.set(FileSystem.FS_DEFAULT_NAME_KEY, "faildel:///");
    conf.setClass("fs.faildel.impl", FakeFileSystem.class, FileSystem.class);
    conf.set(MRJobConfig.TASK_ATTEMPT_ID, attempt);
    conf.setInt(MRJobConfig.APPLICATION_ATTEMPT_ID, 1);
    conf.setInt(FileOutputCommitter.FILEOUTPUTCOMMITTER_ALGORITHM_VERSION, version);
    FileOutputFormat.setOutputPath(job, outDir);
    JobContext jContext = new JobContextImpl(conf, taskID.getJobID());
    TaskAttemptContext tContext = new TaskAttemptContextImpl(conf, taskID);
    FileOutputCommitter committer = new FileOutputCommitter(outDir, tContext);
    // do setup
    committer.setupJob(jContext);
    committer.setupTask(tContext);
    // write output
    TextOutputFormat<?, ?> theOutputFormat = new TextOutputFormat();
    RecordWriter<?, ?> theRecordWriter = theOutputFormat.getRecordWriter(tContext);
    writeOutput(theRecordWriter, tContext);
    // do abort
    Throwable th = null;
    try {
        committer.abortTask(tContext);
    } catch (IOException ie) {
        th = ie;
    }
    assertNotNull(th);
    assertTrue(th instanceof IOException);
    assertTrue(th.getMessage().contains("fake delete failed"));
    Path jtd = committer.getJobAttemptPath(jContext);
    File jobTmpDir = new File(jtd.toUri().getPath());
    Path ttd = committer.getTaskAttemptPath(tContext);
    File taskTmpDir = new File(ttd.toUri().getPath());
    File expectedFile = new File(taskTmpDir, partFile);
    assertTrue(expectedFile + " does not exists", expectedFile.exists());
    th = null;
    try {
        committer.abortJob(jContext, JobStatus.State.FAILED);
    } catch (IOException ie) {
        th = ie;
    }
    assertNotNull(th);
    assertTrue(th instanceof IOException);
    assertTrue(th.getMessage().contains("fake delete failed"));
    assertTrue("job temp dir does not exists", jobTmpDir.exists());
    FileUtil.fullyDelete(new File(outDir.toString()));
}

Example 37 with TaskAttemptContextImpl

use of org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl in project hadoop by apache.

the class TestCombineFileInputFormat method testReinit.

@Test
public void testReinit() throws Exception {
    // Test that a split containing multiple files works correctly,
    // with the child RecordReader getting its initialize() method
    // called a second time.
    TaskAttemptID taskId = new TaskAttemptID("jt", 0, TaskType.MAP, 0, 0);
    Configuration conf = new Configuration();
    TaskAttemptContext context = new TaskAttemptContextImpl(conf, taskId);
    // This will create a CombineFileRecordReader that itself contains a
    // DummyRecordReader.
    InputFormat inputFormat = new ChildRRInputFormat();
    Path[] files = { new Path("file1"), new Path("file2") };
    long[] lengths = { 1, 1 };
    CombineFileSplit split = new CombineFileSplit(files, lengths);
    RecordReader rr = inputFormat.createRecordReader(split, context);
    assertTrue("Unexpected RR type!", rr instanceof CombineFileRecordReader);
    // first initialize() call comes from MapTask. We'll do it here.
    rr.initialize(split, context);
    // First value is first filename.
    assertTrue(rr.nextKeyValue());
    assertEquals("file1", rr.getCurrentValue().toString());
    // The inner RR will return false, because it only emits one (k, v) pair.
    // But there's another sub-split to process. This returns true to us.
    assertTrue(rr.nextKeyValue());
    // And the 2nd rr will have its initialize method called correctly.
    assertEquals("file2", rr.getCurrentValue().toString());
    // But after both child RR's have returned their singleton (k, v), this
    // should also return false.
    assertFalse(rr.nextKeyValue());
}

Example 38 with TaskAttemptContextImpl

use of org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl in project hadoop by apache.

the class TestCombineFileInputFormat method testRecordReaderInit.

@Test
public void testRecordReaderInit() throws InterruptedException, IOException {
    // Test that we properly initialize the child recordreader when
    // CombineFileInputFormat and CombineFileRecordReader are used.
    TaskAttemptID taskId = new TaskAttemptID("jt", 0, TaskType.MAP, 0, 0);
    Configuration conf1 = new Configuration();
    conf1.set(DUMMY_KEY, "STATE1");
    TaskAttemptContext context1 = new TaskAttemptContextImpl(conf1, taskId);
    // This will create a CombineFileRecordReader that itself contains a
    // DummyRecordReader.
    InputFormat inputFormat = new ChildRRInputFormat();
    Path[] files = { new Path("file1") };
    long[] lengths = { 1 };
    CombineFileSplit split = new CombineFileSplit(files, lengths);
    RecordReader rr = inputFormat.createRecordReader(split, context1);
    assertTrue("Unexpected RR type!", rr instanceof CombineFileRecordReader);
    // Verify that the initial configuration is the one being used.
    // Right after construction the dummy key should have value "STATE1"
    assertEquals("Invalid initial dummy key value", "STATE1", rr.getCurrentKey().toString());
    // Switch the active context for the RecordReader...
    Configuration conf2 = new Configuration();
    conf2.set(DUMMY_KEY, "STATE2");
    TaskAttemptContext context2 = new TaskAttemptContextImpl(conf2, taskId);
    rr.initialize(split, context2);
    // And verify that the new context is updated into the child record reader.
    assertEquals("Invalid secondary dummy key value", "STATE2", rr.getCurrentKey().toString());
}

Example 39 with TaskAttemptContextImpl

use of org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl in project hadoop by apache.

the class TestMRCJCFileOutputCommitter method testEmptyOutput.

public void testEmptyOutput() throws Exception {
    Job job = Job.getInstance();
    FileOutputFormat.setOutputPath(job, outDir);
    Configuration conf = job.getConfiguration();
    conf.set(MRJobConfig.TASK_ATTEMPT_ID, attempt);
    JobContext jContext = new JobContextImpl(conf, taskID.getJobID());
    TaskAttemptContext tContext = new TaskAttemptContextImpl(conf, taskID);
    FileOutputCommitter committer = new FileOutputCommitter(outDir, tContext);
    // setup
    committer.setupJob(jContext);
    committer.setupTask(tContext);
    // Do not write any output
    // do commit
    committer.commitTask(tContext);
    committer.commitJob(jContext);
    FileUtil.fullyDelete(new File(outDir.toString()));
}

Example 40 with TaskAttemptContextImpl

use of org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl in project hadoop by apache.

the class TestMRCJCFileOutputCommitter method testFailAbort.

@SuppressWarnings("unchecked")
public void testFailAbort() throws IOException, InterruptedException {
    Job job = Job.getInstance();
    Configuration conf = job.getConfiguration();
    conf.set(FileSystem.FS_DEFAULT_NAME_KEY, "faildel:///");
    conf.setClass("fs.faildel.impl", FakeFileSystem.class, FileSystem.class);
    conf.set(MRJobConfig.TASK_ATTEMPT_ID, attempt);
    FileOutputFormat.setOutputPath(job, outDir);
    JobContext jContext = new JobContextImpl(conf, taskID.getJobID());
    TaskAttemptContext tContext = new TaskAttemptContextImpl(conf, taskID);
    FileOutputCommitter committer = new FileOutputCommitter(outDir, tContext);
    // do setup
    committer.setupJob(jContext);
    committer.setupTask(tContext);
    // write output
    TextOutputFormat<?, ?> theOutputFormat = new TextOutputFormat();
    RecordWriter<?, ?> theRecordWriter = theOutputFormat.getRecordWriter(tContext);
    writeOutput(theRecordWriter, tContext);
    // do abort
    Throwable th = null;
    try {
        committer.abortTask(tContext);
    } catch (IOException ie) {
        th = ie;
    }
    assertNotNull(th);
    assertTrue(th instanceof IOException);
    assertTrue(th.getMessage().contains("fake delete failed"));
    //Path taskBaseDirName = committer.getTaskAttemptBaseDirName(tContext);
    File jobTmpDir = new File(committer.getJobAttemptPath(jContext).toUri().getPath());
    File taskTmpDir = new File(committer.getTaskAttemptPath(tContext).toUri().getPath());
    File expectedFile = new File(taskTmpDir, partFile);
    assertTrue(expectedFile + " does not exists", expectedFile.exists());
    th = null;
    try {
        committer.abortJob(jContext, JobStatus.State.FAILED);
    } catch (IOException ie) {
        th = ie;
    }
    assertNotNull(th);
    assertTrue(th instanceof IOException);
    assertTrue(th.getMessage().contains("fake delete failed"));
    assertTrue("job temp dir does not exists", jobTmpDir.exists());
    FileUtil.fullyDelete(new File(outDir.toString()));
}