Examples with JobContext org.apache.hadoop.mapreduce.JobContext used on opensource projects

Example 56 with JobContext

use of org.apache.hadoop.mapreduce.JobContext 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()));
}

Example 57 with JobContext

use of org.apache.hadoop.mapreduce.JobContext in project hadoop by apache.

the class TestMRCJCFileInputFormat method testLastInputSplitExceedingSplitBoundary.

@Test
@SuppressWarnings({ "rawtypes", "unchecked" })
public void testLastInputSplitExceedingSplitBoundary() throws Exception {
    FileInputFormat fif = new FileInputFormatForTest(1027l * 1024 * 1024, 128l * 1024 * 1024);
    Configuration conf = new Configuration();
    JobContext jobContext = mock(JobContext.class);
    when(jobContext.getConfiguration()).thenReturn(conf);
    List<InputSplit> splits = fif.getSplits(jobContext);
    assertEquals(8, splits.size());
    for (int i = 0; i < splits.size(); i++) {
        InputSplit split = splits.get(i);
        assertEquals(("host" + i), split.getLocations()[0]);
    }
}

Example 58 with JobContext

use of org.apache.hadoop.mapreduce.JobContext in project hadoop by apache.

the class TestMRCJCFileInputFormat method testLastInputSplitSingleSplit.

@Test
@SuppressWarnings({ "rawtypes", "unchecked" })
public void testLastInputSplitSingleSplit() throws Exception {
    FileInputFormat fif = new FileInputFormatForTest(100l * 1024 * 1024, 128l * 1024 * 1024);
    Configuration conf = new Configuration();
    JobContext jobContext = mock(JobContext.class);
    when(jobContext.getConfiguration()).thenReturn(conf);
    List<InputSplit> splits = fif.getSplits(jobContext);
    assertEquals(1, splits.size());
    for (int i = 0; i < splits.size(); i++) {
        InputSplit split = splits.get(i);
        assertEquals(("host" + i), split.getLocations()[0]);
    }
}

Example 59 with JobContext

use of org.apache.hadoop.mapreduce.JobContext in project carbondata by apache.

the class CarbonInputFormat method getTableBlockInfo.

/**
   * Below method will be used to get the table block info
   *
   * @param job       job context
   * @param segmentId number of segment id
   * @return list of table block
   * @throws IOException
   */
private List<TableBlockInfo> getTableBlockInfo(JobContext job, TableSegmentUniqueIdentifier tableSegmentUniqueIdentifier, Set<SegmentTaskIndexStore.TaskBucketHolder> taskKeys, UpdateVO updateDetails, SegmentUpdateStatusManager updateStatusManager, String segmentId, Set<SegmentTaskIndexStore.TaskBucketHolder> validTaskKeys) throws IOException {
    List<TableBlockInfo> tableBlockInfoList = new ArrayList<TableBlockInfo>();
    // get file location of all files of given segment
    JobContext newJob = new JobContextImpl(new Configuration(job.getConfiguration()), job.getJobID());
    newJob.getConfiguration().set(CarbonInputFormat.INPUT_SEGMENT_NUMBERS, tableSegmentUniqueIdentifier.getSegmentId() + "");
    // identify table blocks
    for (InputSplit inputSplit : getSplitsInternal(newJob)) {
        CarbonInputSplit carbonInputSplit = (CarbonInputSplit) inputSplit;
        // then add as TableInfo object.
        if (isValidBlockBasedOnUpdateDetails(taskKeys, carbonInputSplit, updateDetails, updateStatusManager, segmentId, validTaskKeys)) {
            BlockletInfos blockletInfos = new BlockletInfos(carbonInputSplit.getNumberOfBlocklets(), 0, carbonInputSplit.getNumberOfBlocklets());
            tableBlockInfoList.add(new TableBlockInfo(carbonInputSplit.getPath().toString(), carbonInputSplit.getStart(), tableSegmentUniqueIdentifier.getSegmentId(), carbonInputSplit.getLocations(), carbonInputSplit.getLength(), blockletInfos, carbonInputSplit.getVersion(), carbonInputSplit.getBlockStorageIdMap()));
        }
    }
    return tableBlockInfoList;
}

Example 60 with JobContext

use of org.apache.hadoop.mapreduce.JobContext in project jena by apache.

the class AbstractBlankNodeTests method blank_node_identity_02.

/**
     * Test that starts with two blank nodes in two different files and checks
     * that writing them to a single file does not conflate them
     * 
     * @throws IOException
     * @throws InterruptedException
     */
@Test
public void blank_node_identity_02() throws IOException, InterruptedException {
    Assume.assumeTrue("Requires ParserProfile be respected", this.respectsParserProfile());
    Assume.assumeFalse("Requires that Blank Node identity not be preserved", this.preservesBlankNodeIdentity());
    // Temporary files
    File a = File.createTempFile("bnode_identity", getInitialInputExtension());
    File b = File.createTempFile("bnode_identity", getInitialInputExtension());
    File intermediateOutputDir = Files.createTempDirectory("bnode_identity", new FileAttribute[0]).toFile();
    try {
        // Prepare the input data
        // Same blank node but in different files so must be treated as
        // different blank nodes and not converge
        List<T> tuples = new ArrayList<>();
        Node bnode = NodeFactory.createBlankNode();
        Node pred = NodeFactory.createURI("http://example.org/predicate");
        tuples.add(createTuple(bnode, pred, NodeFactory.createLiteral("first")));
        writeTuples(a, tuples);
        tuples.clear();
        tuples.add(createTuple(bnode, pred, NodeFactory.createLiteral("second")));
        writeTuples(b, tuples);
        // Set up fake job which will process the two files
        Configuration config = new Configuration(true);
        InputFormat<LongWritable, TValue> inputFormat = createInitialInputFormat();
        Job job = Job.getInstance(config);
        job.setInputFormatClass(inputFormat.getClass());
        NLineInputFormat.setNumLinesPerSplit(job, 100);
        FileInputFormat.setInputPaths(job, new Path(a.getAbsolutePath()), new Path(b.getAbsolutePath()));
        FileOutputFormat.setOutputPath(job, new Path(intermediateOutputDir.getAbsolutePath()));
        JobContext context = new JobContextImpl(job.getConfiguration(), job.getJobID());
        // Get the splits
        List<InputSplit> splits = inputFormat.getSplits(context);
        Assert.assertEquals(2, splits.size());
        // Prepare the output writing - putting all output to a single file
        OutputFormat<LongWritable, TValue> outputFormat = createIntermediateOutputFormat();
        TaskAttemptContext outputTaskContext = new TaskAttemptContextImpl(job.getConfiguration(), createAttemptID(1, 2, 1));
        RecordWriter<LongWritable, TValue> writer = outputFormat.getRecordWriter(outputTaskContext);
        for (InputSplit split : splits) {
            // Initialize the input reading
            TaskAttemptContext inputTaskContext = new TaskAttemptContextImpl(job.getConfiguration(), createAttemptID(1, 1, 1));
            RecordReader<LongWritable, TValue> reader = inputFormat.createRecordReader(split, inputTaskContext);
            reader.initialize(split, inputTaskContext);
            // output
            while (reader.nextKeyValue()) {
                writer.write(reader.getCurrentKey(), reader.getCurrentValue());
            }
        }
        writer.close(outputTaskContext);
        // Promote outputs from temporary status
        promoteInputs(intermediateOutputDir);
        // Now we need to create a subsequent job that reads the
        // intermediate outputs
        // The Blank nodes should have been given separate identities so we
        // should not be conflating them, this is the opposite problem to
        // that described in JENA-820
        LOGGER.debug("Intermediate output directory is {}", intermediateOutputDir.getAbsolutePath());
        job = Job.getInstance(config);
        inputFormat = createIntermediateInputFormat();
        job.setInputFormatClass(inputFormat.getClass());
        NLineInputFormat.setNumLinesPerSplit(job, 100);
        FileInputFormat.setInputPaths(job, new Path(intermediateOutputDir.getAbsolutePath()));
        context = new JobContextImpl(job.getConfiguration(), job.getJobID());
        // Get the splits
        splits = inputFormat.getSplits(context);
        Assert.assertEquals(1, splits.size());
        // Expect to end up with a single blank node
        Set<Node> nodes = new HashSet<Node>();
        for (InputSplit split : splits) {
            TaskAttemptContext inputTaskContext = new TaskAttemptContextImpl(job.getConfiguration(), new TaskAttemptID());
            RecordReader<LongWritable, TValue> reader = inputFormat.createRecordReader(split, inputTaskContext);
            reader.initialize(split, inputTaskContext);
            while (reader.nextKeyValue()) {
                nodes.add(getSubject(reader.getCurrentValue().get()));
            }
        }
        // Nodes must not diverge
        Assert.assertEquals(2, nodes.size());
    } finally {
        a.delete();
        b.delete();
        deleteDirectory(intermediateOutputDir);
    }
}