Search in sources :

Example 6 with KafkaPartitions

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

the class KafkaSupervisorTest method testBadMetadataOffsets.

@Test(expected = ISE.class)
public void testBadMetadataOffsets() throws Exception {
    supervisor = getSupervisor(1, 1, true, "PT1H", null);
    addSomeEvents(1);
    expect(taskMaster.getTaskRunner()).andReturn(Optional.<TaskRunner>absent()).anyTimes();
    expect(taskStorage.getActiveTasks()).andReturn(ImmutableList.<Task>of()).anyTimes();
    expect(indexerMetadataStorageCoordinator.getDataSourceMetadata(DATASOURCE)).andReturn(new KafkaDataSourceMetadata(new KafkaPartitions(KAFKA_TOPIC, ImmutableMap.of(0, 10L, 1, 20L, 2, 30L)))).anyTimes();
    replayAll();
    supervisor.start();
    supervisor.runInternal();
}
Also used : RealtimeIndexTask(io.druid.indexing.common.task.RealtimeIndexTask) Task(io.druid.indexing.common.task.Task) KafkaIndexTask(io.druid.indexing.kafka.KafkaIndexTask) KafkaPartitions(io.druid.indexing.kafka.KafkaPartitions) KafkaDataSourceMetadata(io.druid.indexing.kafka.KafkaDataSourceMetadata) TaskRunner(io.druid.indexing.overlord.TaskRunner) Test(org.junit.Test)

Example 7 with KafkaPartitions

use of io.druid.indexing.kafka.KafkaPartitions 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!");
        }
    }
}
Also used : TaskResource(io.druid.indexing.common.task.TaskResource) ConcurrentHashMap(java.util.concurrent.ConcurrentHashMap) HashMap(java.util.HashMap) KafkaPartitions(io.druid.indexing.kafka.KafkaPartitions) EntryExistsException(io.druid.metadata.EntryExistsException) DateTime(org.joda.time.DateTime) KafkaIndexTask(io.druid.indexing.kafka.KafkaIndexTask) KafkaIOConfig(io.druid.indexing.kafka.KafkaIOConfig) TaskQueue(io.druid.indexing.overlord.TaskQueue)

Example 8 with KafkaPartitions

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

the class KafkaSupervisorTest method testStopGracefully.

@Test
public void testStopGracefully() throws Exception {
    final TaskLocation location1 = new TaskLocation("testHost", 1234);
    final TaskLocation location2 = new TaskLocation("testHost2", 145);
    final DateTime startTime = new DateTime();
    supervisor = getSupervisor(2, 1, true, "PT1H", null);
    addSomeEvents(1);
    Task id1 = 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);
    Task id2 = createKafkaIndexTask("id2", DATASOURCE, "sequenceName-0", new KafkaPartitions("topic", ImmutableMap.of(0, 10L, 1, 20L, 2, 30L)), new KafkaPartitions("topic", ImmutableMap.of(0, Long.MAX_VALUE, 1, Long.MAX_VALUE, 2, Long.MAX_VALUE)), null);
    Task id3 = createKafkaIndexTask("id3", DATASOURCE, "sequenceName-0", new KafkaPartitions("topic", ImmutableMap.of(0, 10L, 1, 20L, 2, 30L)), 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(id1.getId(), null, location1));
    workItems.add(new TestTaskRunnerWorkItem(id2.getId(), null, location2));
    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(id1, id2, id3)).anyTimes();
    expect(taskStorage.getStatus("id1")).andReturn(Optional.of(TaskStatus.running("id1"))).anyTimes();
    expect(taskStorage.getStatus("id2")).andReturn(Optional.of(TaskStatus.running("id2"))).anyTimes();
    expect(taskStorage.getStatus("id3")).andReturn(Optional.of(TaskStatus.running("id3"))).anyTimes();
    expect(taskStorage.getTask("id1")).andReturn(Optional.of(id1)).anyTimes();
    expect(taskStorage.getTask("id2")).andReturn(Optional.of(id2)).anyTimes();
    expect(taskStorage.getTask("id3")).andReturn(Optional.of(id3)).anyTimes();
    expect(indexerMetadataStorageCoordinator.getDataSourceMetadata(DATASOURCE)).andReturn(new KafkaDataSourceMetadata(null)).anyTimes();
    expect(taskClient.getStatusAsync("id1")).andReturn(Futures.immediateFuture(KafkaIndexTask.Status.PUBLISHING));
    expect(taskClient.getStatusAsync("id2")).andReturn(Futures.immediateFuture(KafkaIndexTask.Status.READING));
    expect(taskClient.getStatusAsync("id3")).andReturn(Futures.immediateFuture(KafkaIndexTask.Status.READING));
    expect(taskClient.getStartTimeAsync("id2")).andReturn(Futures.immediateFuture(startTime));
    expect(taskClient.getStartTimeAsync("id3")).andReturn(Futures.immediateFuture(startTime));
    expect(taskClient.getCurrentOffsets("id1", true)).andReturn(ImmutableMap.of(0, 10L, 1, 20L, 2, 30L));
    taskRunner.registerListener(anyObject(TaskRunnerListener.class), anyObject(Executor.class));
    replayAll();
    supervisor.start();
    supervisor.runInternal();
    verifyAll();
    reset(taskRunner, taskClient, taskQueue);
    expect(taskRunner.getRunningTasks()).andReturn(workItems).anyTimes();
    expect(taskClient.pauseAsync("id2")).andReturn(Futures.immediateFuture((Map<Integer, Long>) ImmutableMap.of(0, 15L, 1, 25L, 2, 30L)));
    expect(taskClient.setEndOffsetsAsync("id2", ImmutableMap.of(0, 15L, 1, 25L, 2, 30L), true)).andReturn(Futures.immediateFuture(true));
    taskQueue.shutdown("id3");
    expectLastCall().times(2);
    replay(taskRunner, taskClient, taskQueue);
    supervisor.gracefulShutdownInternal();
    verifyAll();
}
Also used : TaskRunnerListener(io.druid.indexing.overlord.TaskRunnerListener) RealtimeIndexTask(io.druid.indexing.common.task.RealtimeIndexTask) Task(io.druid.indexing.common.task.Task) KafkaIndexTask(io.druid.indexing.kafka.KafkaIndexTask) Executor(java.util.concurrent.Executor) KafkaPartitions(io.druid.indexing.kafka.KafkaPartitions) ArrayList(java.util.ArrayList) Collection(java.util.Collection) KafkaDataSourceMetadata(io.druid.indexing.kafka.KafkaDataSourceMetadata) Map(java.util.Map) ImmutableMap(com.google.common.collect.ImmutableMap) TaskLocation(io.druid.indexing.common.TaskLocation) DateTime(org.joda.time.DateTime) Test(org.junit.Test)

Example 9 with KafkaPartitions

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

the class KafkaSupervisorTest method testResetNoDataSourceMetadata.

@Test
public void testResetNoDataSourceMetadata() throws Exception {
    supervisor = getSupervisor(1, 1, true, "PT1H", null);
    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();
    taskRunner.registerListener(anyObject(TaskRunnerListener.class), anyObject(Executor.class));
    replayAll();
    supervisor.start();
    supervisor.runInternal();
    verifyAll();
    KafkaDataSourceMetadata resetMetadata = new KafkaDataSourceMetadata(new KafkaPartitions(KAFKA_TOPIC, ImmutableMap.of(1, 1000L, 2, 1000L)));
    reset(indexerMetadataStorageCoordinator);
    // no DataSourceMetadata in metadata store
    expect(indexerMetadataStorageCoordinator.getDataSourceMetadata(DATASOURCE)).andReturn(null);
    replay(indexerMetadataStorageCoordinator);
    supervisor.resetInternal(resetMetadata);
    verifyAll();
}
Also used : TaskRunnerListener(io.druid.indexing.overlord.TaskRunnerListener) RealtimeIndexTask(io.druid.indexing.common.task.RealtimeIndexTask) Task(io.druid.indexing.common.task.Task) KafkaIndexTask(io.druid.indexing.kafka.KafkaIndexTask) Executor(java.util.concurrent.Executor) KafkaPartitions(io.druid.indexing.kafka.KafkaPartitions) KafkaDataSourceMetadata(io.druid.indexing.kafka.KafkaDataSourceMetadata) Test(org.junit.Test)

Example 10 with KafkaPartitions

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

the class KafkaSupervisorTest method testDiscoverExistingPublishingAndReadingTask.

@Test
public void testDiscoverExistingPublishingAndReadingTask() throws Exception {
    final TaskLocation location1 = new TaskLocation("testHost", 1234);
    final TaskLocation location2 = new TaskLocation("testHost2", 145);
    final DateTime startTime = new DateTime();
    supervisor = getSupervisor(1, 1, true, "PT1H", null);
    addSomeEvents(1);
    Task id1 = 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);
    Task id2 = createKafkaIndexTask("id2", DATASOURCE, "sequenceName-0", new KafkaPartitions("topic", ImmutableMap.of(0, 10L, 1, 20L, 2, 30L)), 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(id1.getId(), null, location1));
    workItems.add(new TestTaskRunnerWorkItem(id2.getId(), null, location2));
    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(id1, id2)).anyTimes();
    expect(taskStorage.getStatus("id1")).andReturn(Optional.of(TaskStatus.running("id1"))).anyTimes();
    expect(taskStorage.getStatus("id2")).andReturn(Optional.of(TaskStatus.running("id2"))).anyTimes();
    expect(taskStorage.getTask("id1")).andReturn(Optional.of(id1)).anyTimes();
    expect(taskStorage.getTask("id2")).andReturn(Optional.of(id2)).anyTimes();
    expect(indexerMetadataStorageCoordinator.getDataSourceMetadata(DATASOURCE)).andReturn(new KafkaDataSourceMetadata(null)).anyTimes();
    expect(taskClient.getStatusAsync("id1")).andReturn(Futures.immediateFuture(KafkaIndexTask.Status.PUBLISHING));
    expect(taskClient.getStatusAsync("id2")).andReturn(Futures.immediateFuture(KafkaIndexTask.Status.READING));
    expect(taskClient.getStartTimeAsync("id2")).andReturn(Futures.immediateFuture(startTime));
    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(taskClient.getCurrentOffsetsAsync("id2", false)).andReturn(Futures.immediateFuture((Map<Integer, Long>) ImmutableMap.of(0, 40L, 1, 50L, 2, 60L)));
    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(1, payload.getActiveTasks().size());
    Assert.assertEquals(1, payload.getPublishingTasks().size());
    TaskReportData activeReport = payload.getActiveTasks().get(0);
    TaskReportData publishingReport = payload.getPublishingTasks().get(0);
    Assert.assertEquals("id2", activeReport.getId());
    Assert.assertEquals(startTime, activeReport.getStartTime());
    Assert.assertEquals(ImmutableMap.of(0, 10L, 1, 20L, 2, 30L), activeReport.getStartingOffsets());
    Assert.assertEquals(ImmutableMap.of(0, 40L, 1, 50L, 2, 60L), activeReport.getCurrentOffsets());
    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());
}
Also used : TaskRunnerListener(io.druid.indexing.overlord.TaskRunnerListener) RealtimeIndexTask(io.druid.indexing.common.task.RealtimeIndexTask) Task(io.druid.indexing.common.task.Task) KafkaIndexTask(io.druid.indexing.kafka.KafkaIndexTask) KafkaPartitions(io.druid.indexing.kafka.KafkaPartitions) ArrayList(java.util.ArrayList) TaskLocation(io.druid.indexing.common.TaskLocation) DateTime(org.joda.time.DateTime) Executor(java.util.concurrent.Executor) SupervisorReport(io.druid.indexing.overlord.supervisor.SupervisorReport) Collection(java.util.Collection) KafkaDataSourceMetadata(io.druid.indexing.kafka.KafkaDataSourceMetadata) Map(java.util.Map) ImmutableMap(com.google.common.collect.ImmutableMap) Test(org.junit.Test)

Aggregations

KafkaIndexTask (io.druid.indexing.kafka.KafkaIndexTask)13 KafkaPartitions (io.druid.indexing.kafka.KafkaPartitions)13 RealtimeIndexTask (io.druid.indexing.common.task.RealtimeIndexTask)12 Task (io.druid.indexing.common.task.Task)12 KafkaDataSourceMetadata (io.druid.indexing.kafka.KafkaDataSourceMetadata)12 Test (org.junit.Test)12 TaskRunnerListener (io.druid.indexing.overlord.TaskRunnerListener)10 Executor (java.util.concurrent.Executor)10 TaskLocation (io.druid.indexing.common.TaskLocation)5 ArrayList (java.util.ArrayList)5 Collection (java.util.Collection)5 DateTime (org.joda.time.DateTime)5 ImmutableMap (com.google.common.collect.ImmutableMap)4 KafkaIOConfig (io.druid.indexing.kafka.KafkaIOConfig)4 Map (java.util.Map)4 SupervisorReport (io.druid.indexing.overlord.supervisor.SupervisorReport)3 TaskRunner (io.druid.indexing.overlord.TaskRunner)2 EasyMock.anyString (org.easymock.EasyMock.anyString)2 TaskResource (io.druid.indexing.common.task.TaskResource)1 DataSourceMetadata (io.druid.indexing.overlord.DataSourceMetadata)1