Examples with KafkaIndexTask io.druid.indexing.kafka.KafkaIndexTask used on opensource projects

Example 6 with KafkaIndexTask

use of io.druid.indexing.kafka.KafkaIndexTask in project druid by druid-io.

the class KafkaSupervisorTest method testDiscoverExistingPublishingTask.

@Test
public void testDiscoverExistingPublishingTask() throws Exception {
    final TaskLocation location = new TaskLocation("testHost", 1234);
    supervisor = getSupervisor(1, 1, true, "PT1H", null);
    addSomeEvents(1);
    Task task = createKafkaIndexTask("id1", DATASOURCE, "sequenceName-0", new KafkaPartitions("topic", ImmutableMap.of(0, 0L, 1, 0L, 2, 0L)), new KafkaPartitions("topic", ImmutableMap.of(0, Long.MAX_VALUE, 1, Long.MAX_VALUE, 2, Long.MAX_VALUE)), null);
    Collection workItems = new ArrayList<>();
    workItems.add(new TestTaskRunnerWorkItem(task.getId(), null, location));
    Capture<KafkaIndexTask> captured = Capture.newInstance();
    expect(taskMaster.getTaskQueue()).andReturn(Optional.of(taskQueue)).anyTimes();
    expect(taskMaster.getTaskRunner()).andReturn(Optional.of(taskRunner)).anyTimes();
    expect(taskRunner.getRunningTasks()).andReturn(workItems).anyTimes();
    expect(taskStorage.getActiveTasks()).andReturn(ImmutableList.of(task)).anyTimes();
    expect(taskStorage.getStatus("id1")).andReturn(Optional.of(TaskStatus.running("id1"))).anyTimes();
    expect(taskStorage.getTask("id1")).andReturn(Optional.of(task)).anyTimes();
    expect(indexerMetadataStorageCoordinator.getDataSourceMetadata(DATASOURCE)).andReturn(new KafkaDataSourceMetadata(null)).anyTimes();
    expect(taskClient.getStatusAsync("id1")).andReturn(Futures.immediateFuture(KafkaIndexTask.Status.PUBLISHING));
    expect(taskClient.getCurrentOffsetsAsync("id1", false)).andReturn(Futures.immediateFuture((Map<Integer, Long>) ImmutableMap.of(0, 10L, 1, 20L, 2, 30L)));
    expect(taskClient.getCurrentOffsets("id1", true)).andReturn(ImmutableMap.of(0, 10L, 1, 20L, 2, 30L));
    expect(taskQueue.add(capture(captured))).andReturn(true);
    taskRunner.registerListener(anyObject(TaskRunnerListener.class), anyObject(Executor.class));
    replayAll();
    supervisor.start();
    supervisor.runInternal();
    SupervisorReport report = supervisor.getStatus();
    verifyAll();
    Assert.assertEquals(DATASOURCE, report.getId());
    Assert.assertTrue(report.getPayload() instanceof KafkaSupervisorReport.KafkaSupervisorReportPayload);
    KafkaSupervisorReport.KafkaSupervisorReportPayload payload = (KafkaSupervisorReport.KafkaSupervisorReportPayload) report.getPayload();
    Assert.assertEquals(DATASOURCE, payload.getDataSource());
    Assert.assertEquals(3600L, (long) payload.getDurationSeconds());
    Assert.assertEquals(NUM_PARTITIONS, (int) payload.getPartitions());
    Assert.assertEquals(1, (int) payload.getReplicas());
    Assert.assertEquals(KAFKA_TOPIC, payload.getTopic());
    Assert.assertEquals(0, payload.getActiveTasks().size());
    Assert.assertEquals(1, payload.getPublishingTasks().size());
    TaskReportData publishingReport = payload.getPublishingTasks().get(0);
    Assert.assertEquals("id1", publishingReport.getId());
    Assert.assertEquals(ImmutableMap.of(0, 0L, 1, 0L, 2, 0L), publishingReport.getStartingOffsets());
    Assert.assertEquals(ImmutableMap.of(0, 10L, 1, 20L, 2, 30L), publishingReport.getCurrentOffsets());
    KafkaIndexTask capturedTask = captured.getValue();
    Assert.assertEquals(dataSchema, capturedTask.getDataSchema());
    Assert.assertEquals(KafkaTuningConfig.copyOf(tuningConfig), capturedTask.getTuningConfig());
    KafkaIOConfig capturedTaskConfig = capturedTask.getIOConfig();
    Assert.assertEquals(kafkaHost, capturedTaskConfig.getConsumerProperties().get("bootstrap.servers"));
    Assert.assertEquals("myCustomValue", capturedTaskConfig.getConsumerProperties().get("myCustomKey"));
    Assert.assertEquals("sequenceName-0", capturedTaskConfig.getBaseSequenceName());
    Assert.assertTrue("isUseTransaction", capturedTaskConfig.isUseTransaction());
    Assert.assertFalse("pauseAfterRead", capturedTaskConfig.isPauseAfterRead());
    // check that the new task was created with starting offsets matching where the publishing task finished
    Assert.assertEquals(KAFKA_TOPIC, capturedTaskConfig.getStartPartitions().getTopic());
    Assert.assertEquals(10L, (long) capturedTaskConfig.getStartPartitions().getPartitionOffsetMap().get(0));
    Assert.assertEquals(20L, (long) capturedTaskConfig.getStartPartitions().getPartitionOffsetMap().get(1));
    Assert.assertEquals(30L, (long) capturedTaskConfig.getStartPartitions().getPartitionOffsetMap().get(2));
    Assert.assertEquals(KAFKA_TOPIC, capturedTaskConfig.getEndPartitions().getTopic());
    Assert.assertEquals(Long.MAX_VALUE, (long) capturedTaskConfig.getEndPartitions().getPartitionOffsetMap().get(0));
    Assert.assertEquals(Long.MAX_VALUE, (long) capturedTaskConfig.getEndPartitions().getPartitionOffsetMap().get(1));
    Assert.assertEquals(Long.MAX_VALUE, (long) capturedTaskConfig.getEndPartitions().getPartitionOffsetMap().get(2));
}

Example 7 with KafkaIndexTask

use of io.druid.indexing.kafka.KafkaIndexTask in project druid by druid-io.

the class KafkaSupervisorTest method testDiscoverExistingPublishingTaskWithDifferentPartitionAllocation.

@Test
public void testDiscoverExistingPublishingTaskWithDifferentPartitionAllocation() throws Exception {
    final TaskLocation location = new TaskLocation("testHost", 1234);
    supervisor = getSupervisor(1, 1, true, "PT1H", null);
    addSomeEvents(1);
    Task task = createKafkaIndexTask("id1", DATASOURCE, "sequenceName-0", new KafkaPartitions("topic", ImmutableMap.of(0, 0L, 2, 0L)), new KafkaPartitions("topic", ImmutableMap.of(0, Long.MAX_VALUE, 2, Long.MAX_VALUE)), null);
    Collection workItems = new ArrayList<>();
    workItems.add(new TestTaskRunnerWorkItem(task.getId(), null, location));
    Capture<KafkaIndexTask> captured = Capture.newInstance();
    expect(taskMaster.getTaskQueue()).andReturn(Optional.of(taskQueue)).anyTimes();
    expect(taskMaster.getTaskRunner()).andReturn(Optional.of(taskRunner)).anyTimes();
    expect(taskRunner.getRunningTasks()).andReturn(workItems).anyTimes();
    expect(taskStorage.getActiveTasks()).andReturn(ImmutableList.of(task)).anyTimes();
    expect(taskStorage.getStatus("id1")).andReturn(Optional.of(TaskStatus.running("id1"))).anyTimes();
    expect(taskStorage.getTask("id1")).andReturn(Optional.of(task)).anyTimes();
    expect(indexerMetadataStorageCoordinator.getDataSourceMetadata(DATASOURCE)).andReturn(new KafkaDataSourceMetadata(null)).anyTimes();
    expect(taskClient.getStatusAsync("id1")).andReturn(Futures.immediateFuture(KafkaIndexTask.Status.PUBLISHING));
    expect(taskClient.getCurrentOffsetsAsync("id1", false)).andReturn(Futures.immediateFuture((Map<Integer, Long>) ImmutableMap.of(0, 10L, 2, 30L)));
    expect(taskClient.getCurrentOffsets("id1", true)).andReturn(ImmutableMap.of(0, 10L, 2, 30L));
    expect(taskQueue.add(capture(captured))).andReturn(true);
    taskRunner.registerListener(anyObject(TaskRunnerListener.class), anyObject(Executor.class));
    replayAll();
    supervisor.start();
    supervisor.runInternal();
    SupervisorReport report = supervisor.getStatus();
    verifyAll();
    Assert.assertEquals(DATASOURCE, report.getId());
    Assert.assertTrue(report.getPayload() instanceof KafkaSupervisorReport.KafkaSupervisorReportPayload);
    KafkaSupervisorReport.KafkaSupervisorReportPayload payload = (KafkaSupervisorReport.KafkaSupervisorReportPayload) report.getPayload();
    Assert.assertEquals(DATASOURCE, payload.getDataSource());
    Assert.assertEquals(3600L, (long) payload.getDurationSeconds());
    Assert.assertEquals(NUM_PARTITIONS, (int) payload.getPartitions());
    Assert.assertEquals(1, (int) payload.getReplicas());
    Assert.assertEquals(KAFKA_TOPIC, payload.getTopic());
    Assert.assertEquals(0, payload.getActiveTasks().size());
    Assert.assertEquals(1, payload.getPublishingTasks().size());
    TaskReportData publishingReport = payload.getPublishingTasks().get(0);
    Assert.assertEquals("id1", publishingReport.getId());
    Assert.assertEquals(ImmutableMap.of(0, 0L, 2, 0L), publishingReport.getStartingOffsets());
    Assert.assertEquals(ImmutableMap.of(0, 10L, 2, 30L), publishingReport.getCurrentOffsets());
    KafkaIndexTask capturedTask = captured.getValue();
    Assert.assertEquals(dataSchema, capturedTask.getDataSchema());
    Assert.assertEquals(KafkaTuningConfig.copyOf(tuningConfig), capturedTask.getTuningConfig());
    KafkaIOConfig capturedTaskConfig = capturedTask.getIOConfig();
    Assert.assertEquals(kafkaHost, capturedTaskConfig.getConsumerProperties().get("bootstrap.servers"));
    Assert.assertEquals("myCustomValue", capturedTaskConfig.getConsumerProperties().get("myCustomKey"));
    Assert.assertEquals("sequenceName-0", capturedTaskConfig.getBaseSequenceName());
    Assert.assertTrue("isUseTransaction", capturedTaskConfig.isUseTransaction());
    Assert.assertFalse("pauseAfterRead", capturedTaskConfig.isPauseAfterRead());
    // check that the new task was created with starting offsets matching where the publishing task finished
    Assert.assertEquals(KAFKA_TOPIC, capturedTaskConfig.getStartPartitions().getTopic());
    Assert.assertEquals(10L, (long) capturedTaskConfig.getStartPartitions().getPartitionOffsetMap().get(0));
    Assert.assertEquals(0L, (long) capturedTaskConfig.getStartPartitions().getPartitionOffsetMap().get(1));
    Assert.assertEquals(30L, (long) capturedTaskConfig.getStartPartitions().getPartitionOffsetMap().get(2));
    Assert.assertEquals(KAFKA_TOPIC, capturedTaskConfig.getEndPartitions().getTopic());
    Assert.assertEquals(Long.MAX_VALUE, (long) capturedTaskConfig.getEndPartitions().getPartitionOffsetMap().get(0));
    Assert.assertEquals(Long.MAX_VALUE, (long) capturedTaskConfig.getEndPartitions().getPartitionOffsetMap().get(1));
    Assert.assertEquals(Long.MAX_VALUE, (long) capturedTaskConfig.getEndPartitions().getPartitionOffsetMap().get(2));
}

Example 8 with KafkaIndexTask

use of io.druid.indexing.kafka.KafkaIndexTask in project druid by druid-io.

the class KafkaSupervisorTest method testQueueNextTasksOnSuccess.

@Test
public void testQueueNextTasksOnSuccess() throws Exception {
    supervisor = getSupervisor(2, 2, true, "PT1H", null);
    addSomeEvents(1);
    Capture<Task> captured = Capture.newInstance(CaptureType.ALL);
    expect(taskMaster.getTaskQueue()).andReturn(Optional.of(taskQueue)).anyTimes();
    expect(taskMaster.getTaskRunner()).andReturn(Optional.of(taskRunner)).anyTimes();
    expect(taskRunner.getRunningTasks()).andReturn(Collections.EMPTY_LIST).anyTimes();
    expect(taskStorage.getActiveTasks()).andReturn(ImmutableList.<Task>of()).anyTimes();
    expect(taskClient.getStatusAsync(anyString())).andReturn(Futures.immediateFuture(KafkaIndexTask.Status.NOT_STARTED)).anyTimes();
    expect(taskClient.getStartTimeAsync(anyString())).andReturn(Futures.immediateFuture(DateTime.now())).anyTimes();
    expect(indexerMetadataStorageCoordinator.getDataSourceMetadata(DATASOURCE)).andReturn(new KafkaDataSourceMetadata(null)).anyTimes();
    expect(taskQueue.add(capture(captured))).andReturn(true).times(4);
    taskRunner.registerListener(anyObject(TaskRunnerListener.class), anyObject(Executor.class));
    replayAll();
    supervisor.start();
    supervisor.runInternal();
    verifyAll();
    List<Task> tasks = captured.getValues();
    reset(taskStorage);
    expect(taskStorage.getActiveTasks()).andReturn(tasks).anyTimes();
    for (Task task : tasks) {
        expect(taskStorage.getStatus(task.getId())).andReturn(Optional.of(TaskStatus.running(task.getId()))).anyTimes();
        expect(taskStorage.getTask(task.getId())).andReturn(Optional.of(task)).anyTimes();
    }
    replay(taskStorage);
    supervisor.runInternal();
    verifyAll();
    // test that a task succeeding causes a new task to be re-queued with the next offset range and causes any replica
    // tasks to be shutdown
    Capture<Task> newTasksCapture = Capture.newInstance(CaptureType.ALL);
    Capture<String> shutdownTaskIdCapture = Capture.newInstance();
    List<Task> imStillRunning = tasks.subList(1, 4);
    KafkaIndexTask iAmSuccess = (KafkaIndexTask) tasks.get(0);
    reset(taskStorage);
    reset(taskQueue);
    reset(taskClient);
    expect(taskStorage.getActiveTasks()).andReturn(imStillRunning).anyTimes();
    for (Task task : imStillRunning) {
        expect(taskStorage.getStatus(task.getId())).andReturn(Optional.of(TaskStatus.running(task.getId()))).anyTimes();
        expect(taskStorage.getTask(task.getId())).andReturn(Optional.of(task)).anyTimes();
    }
    expect(taskStorage.getStatus(iAmSuccess.getId())).andReturn(Optional.of(TaskStatus.success(iAmSuccess.getId())));
    expect(taskStorage.getTask(iAmSuccess.getId())).andReturn(Optional.of((Task) iAmSuccess)).anyTimes();
    expect(taskQueue.add(capture(newTasksCapture))).andReturn(true).times(2);
    expect(taskClient.stopAsync(capture(shutdownTaskIdCapture), eq(false))).andReturn(Futures.immediateFuture(true));
    replay(taskStorage);
    replay(taskQueue);
    replay(taskClient);
    supervisor.runInternal();
    verifyAll();
    // make sure we killed the right task (sequenceName for replicas are the same)
    Assert.assertTrue(shutdownTaskIdCapture.getValue().contains(iAmSuccess.getIOConfig().getBaseSequenceName()));
}

Example 9 with KafkaIndexTask

use of io.druid.indexing.kafka.KafkaIndexTask in project druid by druid-io.

the class KafkaSupervisorTest method testLatestOffset.

@Test
public /**
   * Test generating the starting offsets from the partition high water marks in Kafka.
   */
void testLatestOffset() throws Exception {
    supervisor = getSupervisor(1, 1, false, "PT1H", null);
    addSomeEvents(1100);
    Capture<KafkaIndexTask> captured = Capture.newInstance();
    expect(taskMaster.getTaskQueue()).andReturn(Optional.of(taskQueue)).anyTimes();
    expect(taskMaster.getTaskRunner()).andReturn(Optional.<TaskRunner>absent()).anyTimes();
    expect(taskStorage.getActiveTasks()).andReturn(ImmutableList.<Task>of()).anyTimes();
    expect(indexerMetadataStorageCoordinator.getDataSourceMetadata(DATASOURCE)).andReturn(new KafkaDataSourceMetadata(null)).anyTimes();
    expect(taskQueue.add(capture(captured))).andReturn(true);
    replayAll();
    supervisor.start();
    supervisor.runInternal();
    verifyAll();
    KafkaIndexTask task = captured.getValue();
    Assert.assertEquals(1100L, (long) task.getIOConfig().getStartPartitions().getPartitionOffsetMap().get(0));
    Assert.assertEquals(1100L, (long) task.getIOConfig().getStartPartitions().getPartitionOffsetMap().get(1));
    Assert.assertEquals(1100L, (long) task.getIOConfig().getStartPartitions().getPartitionOffsetMap().get(2));
}

Example 10 with KafkaIndexTask

use of io.druid.indexing.kafka.KafkaIndexTask in project druid by druid-io.

the class KafkaSupervisor method createKafkaTasksForGroup.

private void createKafkaTasksForGroup(int groupId, int replicas) {
    Map<Integer, Long> startPartitions = taskGroups.get(groupId).partitionOffsets;
    Map<Integer, Long> endPartitions = new HashMap<>();
    for (Integer partition : startPartitions.keySet()) {
        endPartitions.put(partition, Long.MAX_VALUE);
    }
    String sequenceName = generateSequenceName(groupId);
    Map<String, String> consumerProperties = Maps.newHashMap(ioConfig.getConsumerProperties());
    DateTime minimumMessageTime = taskGroups.get(groupId).minimumMessageTime.orNull();
    KafkaIOConfig kafkaIOConfig = new KafkaIOConfig(sequenceName, new KafkaPartitions(ioConfig.getTopic(), startPartitions), new KafkaPartitions(ioConfig.getTopic(), endPartitions), consumerProperties, true, false, minimumMessageTime);
    for (int i = 0; i < replicas; i++) {
        String taskId = Joiner.on("_").join(sequenceName, getRandomId());
        KafkaIndexTask indexTask = new KafkaIndexTask(taskId, new TaskResource(sequenceName, 1), spec.getDataSchema(), taskTuningConfig, kafkaIOConfig, spec.getContext(), null);
        Optional<TaskQueue> taskQueue = taskMaster.getTaskQueue();
        if (taskQueue.isPresent()) {
            try {
                taskQueue.get().add(indexTask);
            } catch (EntryExistsException e) {
                log.error("Tried to add task [%s] but it already exists", indexTask.getId());
            }
        } else {
            log.error("Failed to get task queue because I'm not the leader!");
        }
    }
}