Examples with InputAttemptIdentifier org.apache.tez.runtime.library.common.InputAttemptIdentifier used on opensource projects

Example 41 with InputAttemptIdentifier

use of org.apache.tez.runtime.library.common.InputAttemptIdentifier in project tez by apache.

the class TestShuffleUtils method testFetchStatsLogger.

@Test
public void testFetchStatsLogger() throws Exception {
    Logger activeLogger = mock(Logger.class);
    Logger aggregateLogger = mock(Logger.class);
    FetchStatsLogger logger = new FetchStatsLogger(activeLogger, aggregateLogger);
    InputAttemptIdentifier ident = new InputAttemptIdentifier(1, 1);
    when(activeLogger.isInfoEnabled()).thenReturn(false);
    for (int i = 0; i < 1000; i++) {
        logger.logIndividualFetchComplete(10, 100, 1000, "testType", ident);
    }
    verify(activeLogger, times(0)).info(anyString());
    verify(aggregateLogger, times(1)).info(anyString(), Matchers.<Object[]>anyVararg());
    when(activeLogger.isInfoEnabled()).thenReturn(true);
    for (int i = 0; i < 1000; i++) {
        logger.logIndividualFetchComplete(10, 100, 1000, "testType", ident);
    }
    verify(activeLogger, times(1000)).info(anyString());
    verify(aggregateLogger, times(1)).info(anyString(), Matchers.<Object[]>anyVararg());
}

Example 42 with InputAttemptIdentifier

use of org.apache.tez.runtime.library.common.InputAttemptIdentifier in project tez by apache.

the class TestFetcher method testWithRetry.

@Test(timeout = 5000)
@SuppressWarnings("unchecked")
public void testWithRetry() throws Exception {
    Configuration conf = new TezConfiguration();
    conf.setInt(TezRuntimeConfiguration.TEZ_RUNTIME_SHUFFLE_READ_TIMEOUT, 3000);
    conf.setInt(TezRuntimeConfiguration.TEZ_RUNTIME_SHUFFLE_CONNECT_TIMEOUT, 3000);
    ShuffleScheduler scheduler = mock(ShuffleScheduler.class);
    MergeManager merger = mock(MergeManager.class);
    Shuffle shuffle = mock(Shuffle.class);
    InputContext inputContext = mock(InputContext.class);
    when(inputContext.getCounters()).thenReturn(new TezCounters());
    when(inputContext.getSourceVertexName()).thenReturn("");
    when(inputContext.getApplicationId()).thenReturn(ApplicationId.newInstance(0, 1));
    HttpConnectionParams httpConnectionParams = ShuffleUtils.getHttpConnectionParams(conf);
    final MapHost host = new MapHost(HOST, PORT, 1, 1);
    FetcherOrderedGrouped mockFetcher = new FetcherOrderedGrouped(null, scheduler, merger, shuffle, null, false, 0, null, conf, false, HOST, PORT, "src vertex", host, ioErrsCounter, wrongLengthErrsCounter, badIdErrsCounter, wrongMapErrsCounter, connectionErrsCounter, wrongReduceErrsCounter, APP_ID, DAG_ID, false, false, true, false);
    final FetcherOrderedGrouped fetcher = spy(mockFetcher);
    final List<InputAttemptIdentifier> srcAttempts = Arrays.asList(new InputAttemptIdentifier(0, 1, InputAttemptIdentifier.PATH_PREFIX + "pathComponent_0"), new InputAttemptIdentifier(1, 2, InputAttemptIdentifier.PATH_PREFIX + "pathComponent_1"), new InputAttemptIdentifier(3, 4, InputAttemptIdentifier.PATH_PREFIX + "pathComponent_3"));
    doReturn(srcAttempts).when(scheduler).getMapsForHost(host);
    doReturn(true).when(fetcher).setupConnection(any(MapHost.class), any(Collection.class));
    URL url = ShuffleUtils.constructInputURL("http://" + HOST + ":" + PORT + "/mapOutput?job=job_123&&reduce=1&map=", srcAttempts, false);
    fetcher.httpConnection = new FakeHttpConnection(url, null, "", null);
    doAnswer(new Answer<MapOutput>() {

        @Override
        public MapOutput answer(InvocationOnMock invocation) throws Throwable {
            Object[] args = invocation.getArguments();
            MapOutput mapOutput = mock(MapOutput.class);
            doReturn(MapOutput.Type.MEMORY).when(mapOutput).getType();
            doReturn(args[0]).when(mapOutput).getAttemptIdentifier();
            return mapOutput;
        }
    }).when(merger).reserve(any(InputAttemptIdentifier.class), anyInt(), anyInt(), anyInt());
    // Create read timeout when reading data
    doAnswer(new Answer<Void>() {

        @Override
        public Void answer(InvocationOnMock invocation) throws Throwable {
            // Emulate host down for 4 seconds.
            Thread.sleep(4000);
            doReturn(false).when(fetcher).setupConnection(any(MapHost.class), any(Collection.class));
            // Throw IOException when fetcher tries to connect again to the same node
            throw new FetcherReadTimeoutException("creating fetcher socket read timeout exception");
        }
    }).when(fetcher).copyMapOutput(any(MapHost.class), any(DataInputStream.class), any(InputAttemptIdentifier.class));
    try {
        fetcher.copyFromHost(host);
    } catch (IOException e) {
    // ignore
    }
    // setup connection should be called twice (1 for connect and another for retry)
    verify(fetcher, times(2)).setupConnection(any(MapHost.class), any(Collection.class));
    // since copyMapOutput consistently fails, it should call copyFailed once
    verify(scheduler, times(1)).copyFailed(any(InputAttemptIdentifier.class), any(MapHost.class), anyBoolean(), anyBoolean(), anyBoolean());
    verify(fetcher, times(1)).putBackRemainingMapOutputs(any(MapHost.class));
    verify(scheduler, times(3)).putBackKnownMapOutput(any(MapHost.class), any(InputAttemptIdentifier.class));
    // Verify by stopping the fetcher abruptly
    try {
        // flag to indicate fetcher stopped
        fetcher.stopped = false;
        fetcher.copyFromHost(host);
        verify(fetcher, times(2)).putBackRemainingMapOutputs(any(MapHost.class));
    } catch (IOException e) {
    // ignore
    }
}

Example 43 with InputAttemptIdentifier

use of org.apache.tez.runtime.library.common.InputAttemptIdentifier in project tez by apache.

the class TestFetcher method testSetupLocalDiskFetchAutoReduce.

@Test(timeout = 5000)
public void testSetupLocalDiskFetchAutoReduce() throws Exception {
    Configuration conf = new TezConfiguration();
    ShuffleScheduler scheduler = mock(ShuffleScheduler.class);
    MergeManager merger = mock(MergeManager.class);
    Shuffle shuffle = mock(Shuffle.class);
    InputContext inputContext = mock(InputContext.class);
    when(inputContext.getCounters()).thenReturn(new TezCounters());
    when(inputContext.getSourceVertexName()).thenReturn("");
    MapHost host = new MapHost(HOST, PORT, 1, 2);
    FetcherOrderedGrouped fetcher = new FetcherOrderedGrouped(null, scheduler, merger, shuffle, null, false, 0, null, conf, true, HOST, PORT, "src vertex", host, ioErrsCounter, wrongLengthErrsCounter, badIdErrsCounter, wrongMapErrsCounter, connectionErrsCounter, wrongReduceErrsCounter, APP_ID, DAG_ID, false, false, true, false);
    FetcherOrderedGrouped spyFetcher = spy(fetcher);
    final List<CompositeInputAttemptIdentifier> srcAttempts = Arrays.asList(new CompositeInputAttemptIdentifier(0, 1, InputAttemptIdentifier.PATH_PREFIX + "pathComponent_0", host.getPartitionCount()), new CompositeInputAttemptIdentifier(1, 2, InputAttemptIdentifier.PATH_PREFIX + "pathComponent_1", host.getPartitionCount()), new CompositeInputAttemptIdentifier(2, 3, InputAttemptIdentifier.PATH_PREFIX + "pathComponent_2", host.getPartitionCount()), new CompositeInputAttemptIdentifier(3, 4, InputAttemptIdentifier.PATH_PREFIX + "pathComponent_3", host.getPartitionCount()), new CompositeInputAttemptIdentifier(4, 4, InputAttemptIdentifier.PATH_PREFIX + "pathComponent_4", host.getPartitionCount()));
    final int FIRST_FAILED_ATTEMPT_IDX = 2;
    final int SECOND_FAILED_ATTEMPT_IDX = 4;
    final int[] sucessfulAttemptsIndexes = { 0, 1, 3 };
    doReturn(srcAttempts).when(scheduler).getMapsForHost(host);
    final ConcurrentMap<ShuffleScheduler.PathPartition, InputAttemptIdentifier> pathToIdentifierMap = new ConcurrentHashMap<ShuffleScheduler.PathPartition, InputAttemptIdentifier>();
    for (CompositeInputAttemptIdentifier srcAttempt : srcAttempts) {
        for (int i = 0; i < srcAttempt.getInputIdentifierCount(); i++) {
            ShuffleScheduler.PathPartition pathPartition = new ShuffleScheduler.PathPartition(srcAttempt.getPathComponent(), host.getPartitionId() + i);
            pathToIdentifierMap.put(pathPartition, srcAttempt.expand(i));
        }
    }
    doAnswer(new Answer<InputAttemptIdentifier>() {

        @Override
        public InputAttemptIdentifier answer(InvocationOnMock invocation) throws Throwable {
            Object[] args = invocation.getArguments();
            String path = (String) args[0];
            int reduceId = (int) args[1];
            return pathToIdentifierMap.get(new ShuffleScheduler.PathPartition(path, reduceId));
        }
    }).when(scheduler).getIdentifierForFetchedOutput(any(String.class), any(int.class));
    doAnswer(new Answer<MapOutput>() {

        @Override
        public MapOutput answer(InvocationOnMock invocation) throws Throwable {
            Object[] args = invocation.getArguments();
            MapOutput mapOutput = mock(MapOutput.class);
            doReturn(MapOutput.Type.DISK_DIRECT).when(mapOutput).getType();
            doReturn(args[0]).when(mapOutput).getAttemptIdentifier();
            return mapOutput;
        }
    }).when(spyFetcher).getMapOutputForDirectDiskFetch(any(InputAttemptIdentifier.class), any(Path.class), any(TezIndexRecord.class));
    doAnswer(new Answer<Path>() {

        @Override
        public Path answer(InvocationOnMock invocation) throws Throwable {
            Object[] args = invocation.getArguments();
            return new Path(SHUFFLE_INPUT_FILE_PREFIX + args[0]);
        }
    }).when(spyFetcher).getShuffleInputFileName(anyString(), anyString());
    for (int i = 0; i < host.getPartitionCount(); i++) {
        doAnswer(new Answer<TezIndexRecord>() {

            @Override
            public TezIndexRecord answer(InvocationOnMock invocation) throws Throwable {
                Object[] args = invocation.getArguments();
                String pathComponent = (String) args[0];
                int len = pathComponent.length();
                long p = Long.valueOf(pathComponent.substring(len - 1, len));
                if (pathComponent.equals(srcAttempts.get(FIRST_FAILED_ATTEMPT_IDX).getPathComponent()) || pathComponent.equals(srcAttempts.get(SECOND_FAILED_ATTEMPT_IDX).getPathComponent())) {
                    throw new IOException("Thowing exception to simulate failure case");
                }
                // match with params for copySucceeded below.
                return new TezIndexRecord(p * 10, (p + 1) * 1000, (p + 2) * 100);
            }
        }).when(spyFetcher).getIndexRecord(anyString(), eq(host.getPartitionId() + i));
    }
    doNothing().when(scheduler).copySucceeded(any(InputAttemptIdentifier.class), any(MapHost.class), anyLong(), anyLong(), anyLong(), any(MapOutput.class), anyBoolean());
    doNothing().when(scheduler).putBackKnownMapOutput(host, srcAttempts.get(FIRST_FAILED_ATTEMPT_IDX).expand(0));
    doNothing().when(scheduler).putBackKnownMapOutput(host, srcAttempts.get(SECOND_FAILED_ATTEMPT_IDX).expand(1));
    doNothing().when(scheduler).putBackKnownMapOutput(host, srcAttempts.get(FIRST_FAILED_ATTEMPT_IDX).expand(0));
    doNothing().when(scheduler).putBackKnownMapOutput(host, srcAttempts.get(SECOND_FAILED_ATTEMPT_IDX).expand(1));
    spyFetcher.setupLocalDiskFetch(host);
    // should have exactly 3 success and 1 failure.
    for (int i : sucessfulAttemptsIndexes) {
        for (int j = 0; j < host.getPartitionCount(); j++) {
            verifyCopySucceeded(scheduler, host, srcAttempts, i, j);
        }
    }
    verify(scheduler).copyFailed(srcAttempts.get(FIRST_FAILED_ATTEMPT_IDX).expand(0), host, true, false, true);
    verify(scheduler).copyFailed(srcAttempts.get(FIRST_FAILED_ATTEMPT_IDX).expand(1), host, true, false, true);
    verify(scheduler).copyFailed(srcAttempts.get(SECOND_FAILED_ATTEMPT_IDX).expand(0), host, true, false, true);
    verify(scheduler).copyFailed(srcAttempts.get(SECOND_FAILED_ATTEMPT_IDX).expand(1), host, true, false, true);
    verify(spyFetcher).putBackRemainingMapOutputs(host);
    verify(scheduler).putBackKnownMapOutput(host, srcAttempts.get(FIRST_FAILED_ATTEMPT_IDX));
    verify(scheduler).putBackKnownMapOutput(host, srcAttempts.get(SECOND_FAILED_ATTEMPT_IDX));
    verify(scheduler).putBackKnownMapOutput(host, srcAttempts.get(FIRST_FAILED_ATTEMPT_IDX));
    verify(scheduler).putBackKnownMapOutput(host, srcAttempts.get(SECOND_FAILED_ATTEMPT_IDX));
}

Example 44 with InputAttemptIdentifier

use of org.apache.tez.runtime.library.common.InputAttemptIdentifier in project tez by apache.

the class TestFetcher method testInputsReturnedOnConnectionException.

@Test(timeout = 5000)
public void testInputsReturnedOnConnectionException() throws Exception {
    Configuration conf = new TezConfiguration();
    ShuffleScheduler scheduler = mock(ShuffleScheduler.class);
    MergeManager merger = mock(MergeManager.class);
    Shuffle shuffle = mock(Shuffle.class);
    InputContext inputContext = mock(InputContext.class);
    doReturn(new TezCounters()).when(inputContext).getCounters();
    doReturn("src vertex").when(inputContext).getSourceVertexName();
    MapHost mapHost = new MapHost(HOST, PORT, 0, 1);
    InputAttemptIdentifier inputAttemptIdentifier = new InputAttemptIdentifier(0, 0, "attempt");
    mapHost.addKnownMap(inputAttemptIdentifier);
    List<InputAttemptIdentifier> mapsForHost = Lists.newArrayList(inputAttemptIdentifier);
    doReturn(mapsForHost).when(scheduler).getMapsForHost(mapHost);
    FetcherOrderedGrouped fetcher = new FetcherOrderedGrouped(null, scheduler, merger, shuffle, null, false, 0, null, conf, false, HOST, PORT, "src vertex", mapHost, ioErrsCounter, wrongLengthErrsCounter, badIdErrsCounter, wrongMapErrsCounter, connectionErrsCounter, wrongReduceErrsCounter, APP_ID, DAG_ID, false, false, true, false);
    fetcher.call();
    verify(scheduler).getMapsForHost(mapHost);
    verify(scheduler).freeHost(mapHost);
    verify(scheduler).putBackKnownMapOutput(mapHost, inputAttemptIdentifier);
}

Example 45 with InputAttemptIdentifier

use of org.apache.tez.runtime.library.common.InputAttemptIdentifier in project tez by apache.

the class TestShuffleScheduler method _testReducerHealth_1.

public void _testReducerHealth_1(Configuration conf) throws IOException {
    long startTime = System.currentTimeMillis() - 500000;
    Shuffle shuffle = mock(Shuffle.class);
    final ShuffleSchedulerForTest scheduler = createScheduler(startTime, 320, shuffle, conf);
    int totalProducerNodes = 20;
    // Generate 320 events
    for (int i = 0; i < 320; i++) {
        CompositeInputAttemptIdentifier inputAttemptIdentifier = new CompositeInputAttemptIdentifier(i, 0, "attempt_", 1);
        scheduler.addKnownMapOutput("host" + (i % totalProducerNodes), 10000, i, inputAttemptIdentifier);
    }
    // 100 succeeds
    for (int i = 0; i < 100; i++) {
        InputAttemptIdentifier inputAttemptIdentifier = new InputAttemptIdentifier(i, 0, "attempt_");
        MapOutput mapOutput = MapOutput.createMemoryMapOutput(inputAttemptIdentifier, mock(FetchedInputAllocatorOrderedGrouped.class), 100, false);
        scheduler.copySucceeded(inputAttemptIdentifier, new MapHost("host" + (i % totalProducerNodes), 10000, i, 1), 100, 200, startTime + (i * 100), mapOutput, false);
    }
    // 99 fails
    for (int i = 100; i < 199; i++) {
        InputAttemptIdentifier inputAttemptIdentifier = new InputAttemptIdentifier(i, 0, "attempt_");
        scheduler.copyFailed(inputAttemptIdentifier, new MapHost("host" + (i % totalProducerNodes), 10000, i, 1), false, true, false);
    }
    InputAttemptIdentifier inputAttemptIdentifier = new InputAttemptIdentifier(200, 0, "attempt_");
    // Should fail here and report exception as reducer is not healthy
    scheduler.copyFailed(inputAttemptIdentifier, new MapHost("host" + (200 % totalProducerNodes), 10000, 200, 1), false, true, false);
    int minFailurePerHost = conf.getInt(TezRuntimeConfiguration.TEZ_RUNTIME_SHUFFLE_MIN_FAILURES_PER_HOST, TezRuntimeConfiguration.TEZ_RUNTIME_SHUFFLE_MIN_FAILURES_PER_HOST_DEFAULT);
    if (minFailurePerHost <= 4) {
        // As per test threshold. Should fail & retrigger shuffle
        verify(shuffle, atLeast(0)).reportException(any(Throwable.class));
    } else if (minFailurePerHost > 100) {
        // host failure is so high that this would not retrigger shuffle re-execution
        verify(shuffle, atLeast(1)).reportException(any(Throwable.class));
    }
}