Search in sources :

Example 1 with HadoopTaskContext

use of org.apache.ignite.internal.processors.hadoop.HadoopTaskContext in project ignite by apache.

the class HadoopConcurrentHashMultimapSelftest method testMultiThreaded.

/**
     * @throws Exception if failed.
     */
public void testMultiThreaded() throws Exception {
    GridUnsafeMemory mem = new GridUnsafeMemory(0);
    X.println("___ Started");
    Random rnd = new GridRandom();
    for (int i = 0; i < 20; i++) {
        HadoopJobInfo job = new JobInfo();
        final HadoopTaskContext taskCtx = new TaskContext();
        final HadoopConcurrentHashMultimap m = new HadoopConcurrentHashMultimap(job, mem, 16);
        final ConcurrentMap<Integer, Collection<Integer>> mm = new ConcurrentHashMap<>();
        X.println("___ MT");
        multithreaded(new Callable<Object>() {

            @Override
            public Object call() throws Exception {
                X.println("___ TH in");
                Random rnd = new GridRandom();
                IntWritable key = new IntWritable();
                IntWritable val = new IntWritable();
                HadoopMultimap.Adder a = m.startAdding(taskCtx);
                for (int i = 0; i < 50000; i++) {
                    int k = rnd.nextInt(32000);
                    int v = rnd.nextInt();
                    key.set(k);
                    val.set(v);
                    a.write(key, val);
                    Collection<Integer> list = mm.get(k);
                    if (list == null) {
                        list = new ConcurrentLinkedQueue<>();
                        Collection<Integer> old = mm.putIfAbsent(k, list);
                        if (old != null)
                            list = old;
                    }
                    list.add(v);
                }
                a.close();
                X.println("___ TH out");
                return null;
            }
        }, 3 + rnd.nextInt(27));
        X.println("___ Check: " + m.capacity());
        assertEquals(mm.size(), m.keys());
        assertTrue(m.capacity() > 32000);
        HadoopTaskInput in = m.input(taskCtx);
        while (in.next()) {
            IntWritable key = (IntWritable) in.key();
            Iterator<?> valsIter = in.values();
            Collection<Integer> vals = mm.remove(key.get());
            assertNotNull(vals);
            while (valsIter.hasNext()) {
                IntWritable val = (IntWritable) valsIter.next();
                assertTrue(vals.remove(val.get()));
            }
            assertTrue(vals.isEmpty());
        }
        in.close();
        m.close();
        assertEquals(0, mem.allocatedSize());
    }
}
Also used : HadoopJobInfo(org.apache.ignite.internal.processors.hadoop.HadoopJobInfo) HadoopTaskInput(org.apache.ignite.internal.processors.hadoop.HadoopTaskInput) HadoopTaskContext(org.apache.ignite.internal.processors.hadoop.HadoopTaskContext) IOException(java.io.IOException) GridRandom(org.apache.ignite.internal.util.GridRandom) Random(java.util.Random) GridRandom(org.apache.ignite.internal.util.GridRandom) HadoopJobInfo(org.apache.ignite.internal.processors.hadoop.HadoopJobInfo) HadoopTaskContext(org.apache.ignite.internal.processors.hadoop.HadoopTaskContext) Collection(java.util.Collection) HadoopConcurrentHashMultimap(org.apache.ignite.internal.processors.hadoop.shuffle.collections.HadoopConcurrentHashMultimap) GridUnsafeMemory(org.apache.ignite.internal.util.offheap.unsafe.GridUnsafeMemory) ConcurrentHashMap(java.util.concurrent.ConcurrentHashMap) ConcurrentLinkedQueue(java.util.concurrent.ConcurrentLinkedQueue) IntWritable(org.apache.hadoop.io.IntWritable)

Example 2 with HadoopTaskContext

use of org.apache.ignite.internal.processors.hadoop.HadoopTaskContext in project ignite by apache.

the class HadoopConcurrentHashMultimapSelftest method testMapSimple.

/** */
public void testMapSimple() throws Exception {
    GridUnsafeMemory mem = new GridUnsafeMemory(0);
    //        mem.listen(new GridOffHeapEventListener() {
    //            @Override public void onEvent(GridOffHeapEvent evt) {
    //                if (evt == GridOffHeapEvent.ALLOCATE)
    //                    U.dumpStack();
    //            }
    //        });
    Random rnd = new Random();
    int mapSize = 16 << rnd.nextInt(3);
    HadoopJobInfo job = new JobInfo();
    HadoopTaskContext taskCtx = new TaskContext();
    HadoopConcurrentHashMultimap m = new HadoopConcurrentHashMultimap(job, mem, mapSize);
    HadoopConcurrentHashMultimap.Adder a = m.startAdding(taskCtx);
    Multimap<Integer, Integer> mm = ArrayListMultimap.create();
    Multimap<Integer, Integer> vis = ArrayListMultimap.create();
    for (int i = 0, vals = 4 * mapSize + rnd.nextInt(25); i < vals; i++) {
        int key = rnd.nextInt(mapSize);
        int val = rnd.nextInt();
        a.write(new IntWritable(key), new IntWritable(val));
        mm.put(key, val);
        X.println("k: " + key + " v: " + val);
        a.close();
        check(m, mm, vis, taskCtx);
        a = m.startAdding(taskCtx);
    }
    //        a.add(new IntWritable(10), new IntWritable(2));
    //        mm.put(10, 2);
    //        check(m, mm);
    a.close();
    X.println("Alloc: " + mem.allocatedSize());
    m.close();
    assertEquals(0, mem.allocatedSize());
}
Also used : HadoopJobInfo(org.apache.ignite.internal.processors.hadoop.HadoopJobInfo) HadoopTaskContext(org.apache.ignite.internal.processors.hadoop.HadoopTaskContext) Random(java.util.Random) GridRandom(org.apache.ignite.internal.util.GridRandom) HadoopJobInfo(org.apache.ignite.internal.processors.hadoop.HadoopJobInfo) HadoopTaskContext(org.apache.ignite.internal.processors.hadoop.HadoopTaskContext) HadoopConcurrentHashMultimap(org.apache.ignite.internal.processors.hadoop.shuffle.collections.HadoopConcurrentHashMultimap) GridUnsafeMemory(org.apache.ignite.internal.util.offheap.unsafe.GridUnsafeMemory) IntWritable(org.apache.hadoop.io.IntWritable)

Example 3 with HadoopTaskContext

use of org.apache.ignite.internal.processors.hadoop.HadoopTaskContext in project ignite by apache.

the class HadoopSkipListSelfTest method testMapSimple.

/**
     * @throws Exception On error.
     */
public void testMapSimple() throws Exception {
    GridUnsafeMemory mem = new GridUnsafeMemory(0);
    //        mem.listen(new GridOffHeapEventListener() {
    //            @Override public void onEvent(GridOffHeapEvent evt) {
    //                if (evt == GridOffHeapEvent.ALLOCATE)
    //                    U.dumpStack();
    //            }
    //        });
    Random rnd = new Random();
    int mapSize = 16 << rnd.nextInt(6);
    HadoopJobInfo job = new JobInfo();
    HadoopTaskContext taskCtx = new TaskContext();
    HadoopMultimap m = new HadoopSkipList(job, mem);
    HadoopMultimap.Adder a = m.startAdding(taskCtx);
    Multimap<Integer, Integer> mm = ArrayListMultimap.create();
    Multimap<Integer, Integer> vis = ArrayListMultimap.create();
    for (int i = 0, vals = 4 * mapSize + rnd.nextInt(25); i < vals; i++) {
        int key = rnd.nextInt(mapSize);
        int val = rnd.nextInt();
        a.write(new IntWritable(key), new IntWritable(val));
        mm.put(key, val);
        X.println("k: " + key + " v: " + val);
        a.close();
        check(m, mm, vis, taskCtx);
        a = m.startAdding(taskCtx);
    }
    //        a.add(new IntWritable(10), new IntWritable(2));
    //        mm.put(10, 2);
    //        check(m, mm);
    a.close();
    X.println("Alloc: " + mem.allocatedSize());
    m.close();
    assertEquals(0, mem.allocatedSize());
}
Also used : HadoopJobInfo(org.apache.ignite.internal.processors.hadoop.HadoopJobInfo) HadoopTaskContext(org.apache.ignite.internal.processors.hadoop.HadoopTaskContext) HadoopMultimap(org.apache.ignite.internal.processors.hadoop.shuffle.collections.HadoopMultimap) Random(java.util.Random) GridRandom(org.apache.ignite.internal.util.GridRandom) HadoopJobInfo(org.apache.ignite.internal.processors.hadoop.HadoopJobInfo) HadoopSkipList(org.apache.ignite.internal.processors.hadoop.shuffle.collections.HadoopSkipList) HadoopTaskContext(org.apache.ignite.internal.processors.hadoop.HadoopTaskContext) GridUnsafeMemory(org.apache.ignite.internal.util.offheap.unsafe.GridUnsafeMemory) IntWritable(org.apache.hadoop.io.IntWritable)

Example 4 with HadoopTaskContext

use of org.apache.ignite.internal.processors.hadoop.HadoopTaskContext in project ignite by apache.

the class HadoopV2JobSelfTest method testCustomSerializationApplying.

/**
     * Tests that {@link HadoopJobEx} provides wrapped serializer if it's set in configuration.
     *
     * @throws IgniteCheckedException If fails.
     */
public void testCustomSerializationApplying() throws IgniteCheckedException {
    JobConf cfg = new JobConf();
    cfg.setMapOutputKeyClass(IntWritable.class);
    cfg.setMapOutputValueClass(Text.class);
    cfg.set(CommonConfigurationKeys.IO_SERIALIZATIONS_KEY, CustomSerialization.class.getName());
    HadoopDefaultJobInfo info = createJobInfo(cfg);
    final UUID uuid = UUID.randomUUID();
    HadoopJobId id = new HadoopJobId(uuid, 1);
    HadoopJobEx job = info.createJob(HadoopV2Job.class, id, log, null, new HadoopHelperImpl());
    HadoopTaskContext taskCtx = job.getTaskContext(new HadoopTaskInfo(HadoopTaskType.MAP, null, 0, 0, null));
    HadoopSerialization ser = taskCtx.keySerialization();
    assertEquals(HadoopSerializationWrapper.class.getName(), ser.getClass().getName());
    DataInput in = new DataInputStream(new ByteArrayInputStream(new byte[0]));
    assertEquals(TEST_SERIALIZED_VALUE, ser.read(in, null).toString());
    ser = taskCtx.valueSerialization();
    assertEquals(HadoopSerializationWrapper.class.getName(), ser.getClass().getName());
    assertEquals(TEST_SERIALIZED_VALUE, ser.read(in, null).toString());
}
Also used : HadoopHelperImpl(org.apache.ignite.internal.processors.hadoop.HadoopHelperImpl) DataInputStream(java.io.DataInputStream) HadoopJobId(org.apache.ignite.internal.processors.hadoop.HadoopJobId) DataInput(java.io.DataInput) ByteArrayInputStream(java.io.ByteArrayInputStream) HadoopJobEx(org.apache.ignite.internal.processors.hadoop.HadoopJobEx) HadoopTaskContext(org.apache.ignite.internal.processors.hadoop.HadoopTaskContext) HadoopTaskInfo(org.apache.ignite.internal.processors.hadoop.HadoopTaskInfo) HadoopDefaultJobInfo(org.apache.ignite.internal.processors.hadoop.HadoopDefaultJobInfo) HadoopSerializationWrapper(org.apache.ignite.internal.processors.hadoop.impl.v2.HadoopSerializationWrapper) UUID(java.util.UUID) HadoopSerialization(org.apache.ignite.internal.processors.hadoop.HadoopSerialization) JobConf(org.apache.hadoop.mapred.JobConf)

Example 5 with HadoopTaskContext

use of org.apache.ignite.internal.processors.hadoop.HadoopTaskContext in project ignite by apache.

the class HadoopShuffleJob method onShuffleMessage.

/**
     * @param src Source.
     * @param msg Message.
     * @throws IgniteCheckedException Exception.
     */
public void onShuffleMessage(T src, HadoopShuffleMessage msg) throws IgniteCheckedException {
    assert msg.buffer() != null;
    assert msg.offset() > 0;
    HadoopTaskContext taskCtx = locReducersCtx.get(msg.reducer()).get();
    HadoopPerformanceCounter perfCntr = HadoopPerformanceCounter.getCounter(taskCtx.counters(), null);
    perfCntr.onShuffleMessage(msg.reducer(), U.currentTimeMillis());
    HadoopMultimap map = getOrCreateMap(locMaps, msg.reducer());
    // Add data from message to the map.
    try (HadoopMultimap.Adder adder = map.startAdding(taskCtx)) {
        final GridUnsafeDataInput dataInput = new GridUnsafeDataInput();
        final UnsafeValue val = new UnsafeValue(msg.buffer());
        msg.visit(new HadoopShuffleMessage.Visitor() {

            /** */
            private HadoopMultimap.Key key;

            @Override
            public void onKey(byte[] buf, int off, int len) throws IgniteCheckedException {
                dataInput.bytes(buf, off, off + len);
                key = adder.addKey(dataInput, key);
            }

            @Override
            public void onValue(byte[] buf, int off, int len) {
                val.off = off;
                val.size = len;
                key.add(val);
            }
        });
    }
    if (embedded) {
        // No immediate response.
        if (localShuffleState(src).onShuffleMessage())
            sendFinishResponse(src, msg.jobId());
    } else
        // Response for every message.
        io.apply(src, new HadoopShuffleAck(msg.id(), msg.jobId()));
}
Also used : HadoopMultimap(org.apache.ignite.internal.processors.hadoop.shuffle.collections.HadoopMultimap) HadoopPerformanceCounter(org.apache.ignite.internal.processors.hadoop.counter.HadoopPerformanceCounter) GridUnsafeDataInput(org.apache.ignite.internal.util.io.GridUnsafeDataInput) IgniteCheckedException(org.apache.ignite.IgniteCheckedException) HadoopTaskContext(org.apache.ignite.internal.processors.hadoop.HadoopTaskContext)

Aggregations

HadoopTaskContext (org.apache.ignite.internal.processors.hadoop.HadoopTaskContext)11 Random (java.util.Random)5 IntWritable (org.apache.hadoop.io.IntWritable)5 HadoopMultimap (org.apache.ignite.internal.processors.hadoop.shuffle.collections.HadoopMultimap)5 GridUnsafeMemory (org.apache.ignite.internal.util.offheap.unsafe.GridUnsafeMemory)5 HadoopJobInfo (org.apache.ignite.internal.processors.hadoop.HadoopJobInfo)4 GridRandom (org.apache.ignite.internal.util.GridRandom)4 ByteArrayInputStream (java.io.ByteArrayInputStream)2 DataInputStream (java.io.DataInputStream)2 IOException (java.io.IOException)2 Collection (java.util.Collection)2 ConcurrentHashMap (java.util.concurrent.ConcurrentHashMap)2 ConcurrentLinkedQueue (java.util.concurrent.ConcurrentLinkedQueue)2 IgniteCheckedException (org.apache.ignite.IgniteCheckedException)2 HadoopSerialization (org.apache.ignite.internal.processors.hadoop.HadoopSerialization)2 HadoopTaskInfo (org.apache.ignite.internal.processors.hadoop.HadoopTaskInfo)2 HadoopTaskInput (org.apache.ignite.internal.processors.hadoop.HadoopTaskInput)2 HadoopPerformanceCounter (org.apache.ignite.internal.processors.hadoop.counter.HadoopPerformanceCounter)2 HadoopConcurrentHashMultimap (org.apache.ignite.internal.processors.hadoop.shuffle.collections.HadoopConcurrentHashMultimap)2 HadoopSkipList (org.apache.ignite.internal.processors.hadoop.shuffle.collections.HadoopSkipList)2