Examples with DataSchema org.apache.druid.segment.indexing.DataSchema used on opensource projects

Example 36 with DataSchema

use of org.apache.druid.segment.indexing.DataSchema in project druid by druid-io.

the class KafkaSupervisorTest method getDataSchema.

private static DataSchema getDataSchema(String dataSource) {
    List<DimensionSchema> dimensions = new ArrayList<>();
    dimensions.add(StringDimensionSchema.create("dim1"));
    dimensions.add(StringDimensionSchema.create("dim2"));
    return new DataSchema(dataSource, new TimestampSpec("timestamp", "iso", null), new DimensionsSpec(dimensions), new AggregatorFactory[] { new CountAggregatorFactory("rows") }, new UniformGranularitySpec(Granularities.HOUR, Granularities.NONE, ImmutableList.of()), null);
}

Example 37 with DataSchema

use of org.apache.druid.segment.indexing.DataSchema in project druid by druid-io.

the class KinesisSupervisorTest method testIsTaskCurrent.

@Test
public void testIsTaskCurrent() {
    DateTime minMessageTime = DateTimes.nowUtc();
    DateTime maxMessageTime = DateTimes.nowUtc().plus(10000);
    KinesisSupervisor supervisor = getSupervisor(1, 1, true, "PT1H", new Period("P1D"), new Period("P1D"), false, 42, 42, dataSchema, tuningConfig);
    supervisor.addTaskGroupToActivelyReadingTaskGroup(42, ImmutableMap.of(SHARD_ID1, "3"), Optional.of(minMessageTime), Optional.of(maxMessageTime), ImmutableSet.of("id1", "id2", "id3", "id4"), ImmutableSet.of());
    DataSchema modifiedDataSchema = getDataSchema("some other datasource");
    KinesisSupervisorTuningConfig modifiedTuningConfig = new KinesisSupervisorTuningConfig(null, 1000, null, null, 50000, null, new Period("P1Y"), new File("/test"), null, null, null, false, null, null, null, null, numThreads, TEST_CHAT_THREADS, TEST_CHAT_RETRIES, TEST_HTTP_TIMEOUT, TEST_SHUTDOWN_TIMEOUT, null, null, null, 5000, null, null, null, null, // This property is different from tuningConfig
    42, null, null, null, null);
    KinesisIndexTask taskFromStorage = createKinesisIndexTask("id1", 0, new SeekableStreamStartSequenceNumbers<>("stream", ImmutableMap.of(SHARD_ID1, "3"), ImmutableSet.of()), new SeekableStreamEndSequenceNumbers<>("stream", ImmutableMap.of(SHARD_ID1, KinesisSequenceNumber.NO_END_SEQUENCE_NUMBER)), minMessageTime, maxMessageTime, dataSchema);
    KinesisIndexTask taskFromStorageMismatchedDataSchema = createKinesisIndexTask("id2", 0, new SeekableStreamStartSequenceNumbers<>("stream", ImmutableMap.of(SHARD_ID1, "3"), ImmutableSet.of()), new SeekableStreamEndSequenceNumbers<>("stream", ImmutableMap.of(SHARD_ID1, KinesisSequenceNumber.NO_END_SEQUENCE_NUMBER)), minMessageTime, maxMessageTime, modifiedDataSchema);
    KinesisIndexTask taskFromStorageMismatchedTuningConfig = createKinesisIndexTask("id3", 0, new SeekableStreamStartSequenceNumbers<>("stream", ImmutableMap.of(SHARD_ID1, "3"), ImmutableSet.of()), new SeekableStreamEndSequenceNumbers<>("stream", ImmutableMap.of(SHARD_ID1, KinesisSequenceNumber.NO_END_SEQUENCE_NUMBER)), minMessageTime, maxMessageTime, dataSchema, modifiedTuningConfig);
    KinesisIndexTask taskFromStorageMismatchedPartitionsWithTaskGroup = createKinesisIndexTask("id4", 0, new SeekableStreamStartSequenceNumbers<>("stream", ImmutableMap.of(SHARD_ID1, // this is the mismatch
    "4"), ImmutableSet.of()), new SeekableStreamEndSequenceNumbers<>("stream", ImmutableMap.of(SHARD_ID1, KinesisSequenceNumber.NO_END_SEQUENCE_NUMBER)), minMessageTime, maxMessageTime, dataSchema);
    EasyMock.expect(taskStorage.getTask("id1")).andReturn(Optional.of(taskFromStorage)).once();
    EasyMock.expect(taskStorage.getTask("id2")).andReturn(Optional.of(taskFromStorageMismatchedDataSchema)).once();
    EasyMock.expect(taskStorage.getTask("id3")).andReturn(Optional.of(taskFromStorageMismatchedTuningConfig)).once();
    EasyMock.expect(taskStorage.getTask("id4")).andReturn(Optional.of(taskFromStorageMismatchedPartitionsWithTaskGroup)).once();
    replayAll();
    Assert.assertTrue(supervisor.isTaskCurrent(42, "id1"));
    Assert.assertFalse(supervisor.isTaskCurrent(42, "id2"));
    Assert.assertFalse(supervisor.isTaskCurrent(42, "id3"));
    Assert.assertFalse(supervisor.isTaskCurrent(42, "id4"));
    verifyAll();
}

Example 38 with DataSchema

use of org.apache.druid.segment.indexing.DataSchema in project hive by apache.

the class TestDruidRecordWriter method testWrite.

// Test is failing due to Guava dependency, Druid 0.13.0 should have less dependency on Guava
@Ignore
@Test
public void testWrite() throws IOException, SegmentLoadingException {
    final String dataSourceName = "testDataSource";
    final File segmentOutputDir = temporaryFolder.newFolder();
    final File workingDir = temporaryFolder.newFolder();
    Configuration config = new Configuration();
    final InputRowParser inputRowParser = new MapInputRowParser(new TimeAndDimsParseSpec(new TimestampSpec(DruidConstants.DEFAULT_TIMESTAMP_COLUMN, "auto", null), new DimensionsSpec(ImmutableList.of(new StringDimensionSchema("host")), null, null)));
    final Map<String, Object> parserMap = objectMapper.convertValue(inputRowParser, new TypeReference<Map<String, Object>>() {
    });
    DataSchema dataSchema = new DataSchema(dataSourceName, parserMap, new AggregatorFactory[] { new LongSumAggregatorFactory("visited_sum", "visited_sum"), new HyperUniquesAggregatorFactory("unique_hosts", "unique_hosts") }, new UniformGranularitySpec(Granularities.DAY, Granularities.NONE, ImmutableList.of(INTERVAL_FULL)), null, objectMapper);
    IndexSpec indexSpec = new IndexSpec(new RoaringBitmapSerdeFactory(true), null, null, null);
    RealtimeTuningConfig tuningConfig = new RealtimeTuningConfig(null, null, null, null, temporaryFolder.newFolder(), null, null, null, null, indexSpec, null, null, 0, 0, null, null, 0L, null, null);
    LocalFileSystem localFileSystem = FileSystem.getLocal(config);
    DataSegmentPusher dataSegmentPusher = new LocalDataSegmentPusher(new LocalDataSegmentPusherConfig() {

        @Override
        public File getStorageDirectory() {
            return segmentOutputDir;
        }
    });
    Path segmentDescriptorPath = new Path(workingDir.getAbsolutePath(), DruidStorageHandler.SEGMENTS_DESCRIPTOR_DIR_NAME);
    DruidRecordWriter druidRecordWriter = new DruidRecordWriter(dataSchema, tuningConfig, dataSegmentPusher, 20, segmentDescriptorPath, localFileSystem);
    List<DruidWritable> druidWritables = expectedRows.stream().map(input -> new DruidWritable(ImmutableMap.<String, Object>builder().putAll(input).put(Constants.DRUID_TIMESTAMP_GRANULARITY_COL_NAME, Granularities.DAY.bucketStart(new DateTime((long) input.get(DruidConstants.DEFAULT_TIMESTAMP_COLUMN))).getMillis()).build())).collect(Collectors.toList());
    for (DruidWritable druidWritable : druidWritables) {
        druidRecordWriter.write(druidWritable);
    }
    druidRecordWriter.close(false);
    List<DataSegment> dataSegmentList = DruidStorageHandlerUtils.getCreatedSegments(segmentDescriptorPath, config);
    Assert.assertEquals(1, dataSegmentList.size());
    File tmpUnzippedSegmentDir = temporaryFolder.newFolder();
    new LocalDataSegmentPuller().getSegmentFiles(dataSegmentList.get(0), tmpUnzippedSegmentDir);
    final QueryableIndex queryableIndex = DruidStorageHandlerUtils.INDEX_IO.loadIndex(tmpUnzippedSegmentDir);
    QueryableIndexStorageAdapter adapter = new QueryableIndexStorageAdapter(queryableIndex);
    Firehose firehose = new IngestSegmentFirehose(ImmutableList.of(new WindowedStorageAdapter(adapter, adapter.getInterval())), null, ImmutableList.of("host"), ImmutableList.of("visited_sum", "unique_hosts"), null);
    List<InputRow> rows = Lists.newArrayList();
    while (firehose.hasMore()) {
        rows.add(firehose.nextRow());
    }
    verifyRows(expectedRows, rows);
}

Example 39 with DataSchema

use of org.apache.druid.segment.indexing.DataSchema in project druid by druid-io.

the class MaterializedViewSupervisorTest method testCheckSegmentsAndSubmitTasks.

@Test
public void testCheckSegmentsAndSubmitTasks() throws IOException {
    Set<DataSegment> baseSegments = Sets.newHashSet(new DataSegment("base", Intervals.of("2015-01-02T00Z/2015-01-03T00Z"), "2015-01-03", ImmutableMap.of(), ImmutableList.of("dim1", "dim2"), ImmutableList.of("m1"), new HashBasedNumberedShardSpec(0, 1, 0, 1, null, null, null), 9, 1024));
    indexerMetadataStorageCoordinator.announceHistoricalSegments(baseSegments);
    EasyMock.expect(taskMaster.getTaskQueue()).andReturn(Optional.of(taskQueue)).anyTimes();
    EasyMock.expect(taskMaster.getTaskRunner()).andReturn(Optional.absent()).anyTimes();
    EasyMock.expect(taskStorage.getActiveTasks()).andReturn(ImmutableList.of()).anyTimes();
    EasyMock.expect(taskStorage.getStatus("test_task1")).andReturn(Optional.of(TaskStatus.failure("test_task1", "Dummy task status failure err message"))).anyTimes();
    EasyMock.expect(taskStorage.getStatus("test_task2")).andReturn(Optional.of(TaskStatus.running("test_task2"))).anyTimes();
    EasyMock.replay(taskStorage);
    Pair<Map<Interval, HadoopIndexTask>, Map<Interval, String>> runningTasksPair = supervisor.getRunningTasks();
    Map<Interval, HadoopIndexTask> runningTasks = runningTasksPair.lhs;
    Map<Interval, String> runningVersion = runningTasksPair.rhs;
    DataSchema dataSchema = new DataSchema("test_datasource", null, null, null, TransformSpec.NONE, objectMapper);
    HadoopIOConfig hadoopIOConfig = new HadoopIOConfig(new HashMap<>(), null, null);
    HadoopIngestionSpec spec = new HadoopIngestionSpec(dataSchema, hadoopIOConfig, null);
    HadoopIndexTask task1 = new HadoopIndexTask("test_task1", spec, null, null, null, objectMapper, null, null, null);
    runningTasks.put(Intervals.of("2015-01-01T00Z/2015-01-02T00Z"), task1);
    runningVersion.put(Intervals.of("2015-01-01T00Z/2015-01-02T00Z"), "test_version1");
    HadoopIndexTask task2 = new HadoopIndexTask("test_task2", spec, null, null, null, objectMapper, null, null, null);
    runningTasks.put(Intervals.of("2015-01-02T00Z/2015-01-03T00Z"), task2);
    runningVersion.put(Intervals.of("2015-01-02T00Z/2015-01-03T00Z"), "test_version2");
    supervisor.checkSegmentsAndSubmitTasks();
    Map<Interval, HadoopIndexTask> expectedRunningTasks = new HashMap<>();
    Map<Interval, String> expectedRunningVersion = new HashMap<>();
    expectedRunningTasks.put(Intervals.of("2015-01-02T00Z/2015-01-03T00Z"), task2);
    expectedRunningVersion.put(Intervals.of("2015-01-02T00Z/2015-01-03T00Z"), "test_version2");
    Assert.assertEquals(expectedRunningTasks, runningTasks);
    Assert.assertEquals(expectedRunningVersion, runningVersion);
}

Example 40 with DataSchema

use of org.apache.druid.segment.indexing.DataSchema in project druid by druid-io.

the class IndexTask method generateAndPublishSegments.

/**
 * This method reads input data row by row and adds the read row to a proper segment using {@link BaseAppenderatorDriver}.
 * If there is no segment for the row, a new one is created.  Segments can be published in the middle of reading inputs
 * if {@link DynamicPartitionsSpec} is used and one of below conditions are satisfied.
 *
 * <ul>
 * <li>
 * If the number of rows in a segment exceeds {@link DynamicPartitionsSpec#maxRowsPerSegment}
 * </li>
 * <li>
 * If the number of rows added to {@link BaseAppenderatorDriver} so far exceeds {@link DynamicPartitionsSpec#maxTotalRows}
 * </li>
 * </ul>
 * <p>
 * At the end of this method, all the remaining segments are published.
 *
 * @return the last {@link TaskStatus}
 */
private TaskStatus generateAndPublishSegments(final TaskToolbox toolbox, final DataSchema dataSchema, final InputSource inputSource, final File tmpDir, final PartitionAnalysis partitionAnalysis) throws IOException, InterruptedException {
    final FireDepartment fireDepartmentForMetrics = new FireDepartment(dataSchema, new RealtimeIOConfig(null, null), null);
    FireDepartmentMetrics buildSegmentsFireDepartmentMetrics = fireDepartmentForMetrics.getMetrics();
    if (toolbox.getMonitorScheduler() != null) {
        final TaskRealtimeMetricsMonitor metricsMonitor = TaskRealtimeMetricsMonitorBuilder.build(this, fireDepartmentForMetrics, buildSegmentsMeters);
        toolbox.getMonitorScheduler().addMonitor(metricsMonitor);
    }
    final PartitionsSpec partitionsSpec = partitionAnalysis.getPartitionsSpec();
    final IndexTuningConfig tuningConfig = ingestionSchema.getTuningConfig();
    final long pushTimeout = tuningConfig.getPushTimeout();
    final SegmentAllocatorForBatch segmentAllocator;
    final SequenceNameFunction sequenceNameFunction;
    switch(partitionsSpec.getType()) {
        case HASH:
        case RANGE:
            final SegmentAllocatorForBatch localSegmentAllocator = SegmentAllocators.forNonLinearPartitioning(toolbox, getDataSource(), baseSequenceName, dataSchema.getGranularitySpec(), null, (CompletePartitionAnalysis) partitionAnalysis);
            sequenceNameFunction = localSegmentAllocator.getSequenceNameFunction();
            segmentAllocator = localSegmentAllocator;
            break;
        case LINEAR:
            segmentAllocator = SegmentAllocators.forLinearPartitioning(toolbox, baseSequenceName, null, dataSchema, getTaskLockHelper(), ingestionSchema.getIOConfig().isAppendToExisting(), partitionAnalysis.getPartitionsSpec(), null);
            sequenceNameFunction = segmentAllocator.getSequenceNameFunction();
            break;
        default:
            throw new UOE("[%s] secondary partition type is not supported", partitionsSpec.getType());
    }
    Set<DataSegment> segmentsFoundForDrop = null;
    if (ingestionSchema.getIOConfig().isDropExisting()) {
        segmentsFoundForDrop = getUsedSegmentsWithinInterval(toolbox, getDataSource(), ingestionSchema.getDataSchema().getGranularitySpec().inputIntervals());
    }
    final TransactionalSegmentPublisher publisher = (segmentsToBeOverwritten, segmentsToDrop, segmentsToPublish, commitMetadata) -> toolbox.getTaskActionClient().submit(SegmentTransactionalInsertAction.overwriteAction(segmentsToBeOverwritten, segmentsToDrop, segmentsToPublish));
    String effectiveId = getContextValue(CompactionTask.CTX_KEY_APPENDERATOR_TRACKING_TASK_ID, null);
    if (effectiveId == null) {
        effectiveId = getId();
    }
    final Appenderator appenderator = BatchAppenderators.newAppenderator(effectiveId, toolbox.getAppenderatorsManager(), buildSegmentsFireDepartmentMetrics, toolbox, dataSchema, tuningConfig, buildSegmentsMeters, buildSegmentsParseExceptionHandler, isUseMaxMemoryEstimates());
    boolean exceptionOccurred = false;
    try (final BatchAppenderatorDriver driver = BatchAppenderators.newDriver(appenderator, toolbox, segmentAllocator)) {
        driver.startJob();
        InputSourceProcessor.process(dataSchema, driver, partitionsSpec, inputSource, inputSource.needsFormat() ? getInputFormat(ingestionSchema) : null, tmpDir, sequenceNameFunction, new DefaultIndexTaskInputRowIteratorBuilder(), buildSegmentsMeters, buildSegmentsParseExceptionHandler, pushTimeout);
        // If we use timeChunk lock, then we don't have to specify what segments will be overwritten because
        // it will just overwrite all segments overlapped with the new segments.
        final Set<DataSegment> inputSegments = getTaskLockHelper().isUseSegmentLock() ? getTaskLockHelper().getLockedExistingSegments() : null;
        final boolean storeCompactionState = getContextValue(Tasks.STORE_COMPACTION_STATE_KEY, Tasks.DEFAULT_STORE_COMPACTION_STATE);
        final Function<Set<DataSegment>, Set<DataSegment>> annotateFunction = compactionStateAnnotateFunction(storeCompactionState, toolbox, ingestionSchema);
        // Probably we can publish atomicUpdateGroup along with segments.
        final SegmentsAndCommitMetadata published = awaitPublish(driver.publishAll(inputSegments, segmentsFoundForDrop, publisher, annotateFunction), pushTimeout);
        appenderator.close();
        // for awaitSegmentAvailabilityTimeoutMillis
        if (tuningConfig.getAwaitSegmentAvailabilityTimeoutMillis() > 0 && published != null) {
            ingestionState = IngestionState.SEGMENT_AVAILABILITY_WAIT;
            ArrayList<DataSegment> segmentsToWaitFor = new ArrayList<>(published.getSegments());
            waitForSegmentAvailability(toolbox, segmentsToWaitFor, tuningConfig.getAwaitSegmentAvailabilityTimeoutMillis());
        }
        ingestionState = IngestionState.COMPLETED;
        if (published == null) {
            log.error("Failed to publish segments, aborting!");
            errorMsg = "Failed to publish segments.";
            toolbox.getTaskReportFileWriter().write(getId(), getTaskCompletionReports());
            return TaskStatus.failure(getId(), errorMsg);
        } else {
            log.info("Processed[%,d] events, unparseable[%,d], thrownAway[%,d].", buildSegmentsMeters.getProcessed(), buildSegmentsMeters.getUnparseable(), buildSegmentsMeters.getThrownAway());
            log.info("Published [%s] segments", published.getSegments().size());
            log.debugSegments(published.getSegments(), "Published segments");
            toolbox.getTaskReportFileWriter().write(getId(), getTaskCompletionReports());
            return TaskStatus.success(getId());
        }
    } catch (TimeoutException | ExecutionException e) {
        exceptionOccurred = true;
        throw new RuntimeException(e);
    } catch (Exception e) {
        exceptionOccurred = true;
        throw e;
    } finally {
        if (exceptionOccurred) {
            appenderator.closeNow();
        } else {
            appenderator.close();
        }
    }
}