Examples with DataSourceMetadata org.apache.druid.indexing.overlord.DataSourceMetadata used on opensource projects

Example 1 with DataSourceMetadata

use of org.apache.druid.indexing.overlord.DataSourceMetadata in project druid by druid-io.

the class IndexerSQLMetadataStorageCoordinator method updateDataSourceMetadataWithHandle.

/**
 * Compare-and-swap dataSource metadata in a transaction. This will only modify dataSource metadata if it equals
 * oldCommitMetadata when this function is called (based on T.equals). This method is idempotent in that if
 * the metadata already equals newCommitMetadata, it will return true.
 *
 * @param handle        database handle
 * @param dataSource    druid dataSource
 * @param startMetadata dataSource metadata pre-insert must match this startMetadata according to
 *                      {@link DataSourceMetadata#matches(DataSourceMetadata)}
 * @param endMetadata   dataSource metadata post-insert will have this endMetadata merged in with
 *                      {@link DataSourceMetadata#plus(DataSourceMetadata)}
 *
 * @return SUCCESS if dataSource metadata was updated from matching startMetadata to matching endMetadata, FAILURE or
 * TRY_AGAIN if it definitely was not updated. This guarantee is meant to help
 * {@link #announceHistoricalSegments(Set, Set, DataSourceMetadata, DataSourceMetadata)}
 * achieve its own guarantee.
 *
 * @throws RuntimeException if state is unknown after this call
 */
protected DataStoreMetadataUpdateResult updateDataSourceMetadataWithHandle(final Handle handle, final String dataSource, final DataSourceMetadata startMetadata, final DataSourceMetadata endMetadata) throws IOException {
    Preconditions.checkNotNull(dataSource, "dataSource");
    Preconditions.checkNotNull(startMetadata, "startMetadata");
    Preconditions.checkNotNull(endMetadata, "endMetadata");
    final byte[] oldCommitMetadataBytesFromDb = retrieveDataSourceMetadataWithHandleAsBytes(handle, dataSource);
    final String oldCommitMetadataSha1FromDb;
    final DataSourceMetadata oldCommitMetadataFromDb;
    if (oldCommitMetadataBytesFromDb == null) {
        oldCommitMetadataSha1FromDb = null;
        oldCommitMetadataFromDb = null;
    } else {
        oldCommitMetadataSha1FromDb = BaseEncoding.base16().encode(Hashing.sha1().hashBytes(oldCommitMetadataBytesFromDb).asBytes());
        oldCommitMetadataFromDb = jsonMapper.readValue(oldCommitMetadataBytesFromDb, DataSourceMetadata.class);
    }
    final boolean startMetadataMatchesExisting;
    if (oldCommitMetadataFromDb == null) {
        startMetadataMatchesExisting = startMetadata.isValidStart();
    } else {
        // Checking against the last committed metadata.
        // Converting the last one into start metadata for checking since only the same type of metadata can be matched.
        // Even though kafka/kinesis indexing services use different sequenceNumber types for representing
        // start and end sequenceNumbers, the below conversion is fine because the new start sequenceNumbers are supposed
        // to be same with end sequenceNumbers of the last commit.
        startMetadataMatchesExisting = startMetadata.asStartMetadata().matches(oldCommitMetadataFromDb.asStartMetadata());
    }
    if (!startMetadataMatchesExisting) {
        // Not in the desired start state.
        log.error("Not updating metadata, existing state[%s] in metadata store doesn't match to the new start state[%s].", oldCommitMetadataFromDb, startMetadata);
        return DataStoreMetadataUpdateResult.FAILURE;
    }
    // Only endOffsets should be stored in metadata store
    final DataSourceMetadata newCommitMetadata = oldCommitMetadataFromDb == null ? endMetadata : oldCommitMetadataFromDb.plus(endMetadata);
    final byte[] newCommitMetadataBytes = jsonMapper.writeValueAsBytes(newCommitMetadata);
    final String newCommitMetadataSha1 = BaseEncoding.base16().encode(Hashing.sha1().hashBytes(newCommitMetadataBytes).asBytes());
    final DataStoreMetadataUpdateResult retVal;
    if (oldCommitMetadataBytesFromDb == null) {
        // SELECT -> INSERT can fail due to races; callers must be prepared to retry.
        final int numRows = handle.createStatement(StringUtils.format("INSERT INTO %s (dataSource, created_date, commit_metadata_payload, commit_metadata_sha1) " + "VALUES (:dataSource, :created_date, :commit_metadata_payload, :commit_metadata_sha1)", dbTables.getDataSourceTable())).bind("dataSource", dataSource).bind("created_date", DateTimes.nowUtc().toString()).bind("commit_metadata_payload", newCommitMetadataBytes).bind("commit_metadata_sha1", newCommitMetadataSha1).execute();
        retVal = numRows == 1 ? DataStoreMetadataUpdateResult.SUCCESS : DataStoreMetadataUpdateResult.TRY_AGAIN;
    } else {
        // Expecting a particular old metadata; use the SHA1 in a compare-and-swap UPDATE
        final int numRows = handle.createStatement(StringUtils.format("UPDATE %s SET " + "commit_metadata_payload = :new_commit_metadata_payload, " + "commit_metadata_sha1 = :new_commit_metadata_sha1 " + "WHERE dataSource = :dataSource AND commit_metadata_sha1 = :old_commit_metadata_sha1", dbTables.getDataSourceTable())).bind("dataSource", dataSource).bind("old_commit_metadata_sha1", oldCommitMetadataSha1FromDb).bind("new_commit_metadata_payload", newCommitMetadataBytes).bind("new_commit_metadata_sha1", newCommitMetadataSha1).execute();
        retVal = numRows == 1 ? DataStoreMetadataUpdateResult.SUCCESS : DataStoreMetadataUpdateResult.TRY_AGAIN;
    }
    if (retVal == DataStoreMetadataUpdateResult.SUCCESS) {
        log.info("Updated metadata from[%s] to[%s].", oldCommitMetadataFromDb, newCommitMetadata);
    } else {
        log.info("Not updating metadata, compare-and-swap failure.");
    }
    return retVal;
}

Example 2 with DataSourceMetadata

use of org.apache.druid.indexing.overlord.DataSourceMetadata in project druid by druid-io.

the class SeekableStreamSupervisor method resetInternal.

@VisibleForTesting
public void resetInternal(DataSourceMetadata dataSourceMetadata) {
    if (dataSourceMetadata == null) {
        // Reset everything
        boolean result = indexerMetadataStorageCoordinator.deleteDataSourceMetadata(dataSource);
        log.info("Reset dataSource[%s] - dataSource metadata entry deleted? [%s]", dataSource, result);
        activelyReadingTaskGroups.values().forEach(group -> killTasksInGroup(group, "DataSourceMetadata is not found while reset"));
        activelyReadingTaskGroups.clear();
        partitionGroups.clear();
        partitionOffsets.clear();
    } else {
        if (!checkSourceMetadataMatch(dataSourceMetadata)) {
            throw new IAE("Datasource metadata instance does not match required, found instance of [%s]", dataSourceMetadata.getClass());
        }
        log.info("Reset dataSource[%s] with metadata[%s]", dataSource, dataSourceMetadata);
        // Reset only the partitions in dataSourceMetadata if it has not been reset yet
        @SuppressWarnings("unchecked") final SeekableStreamDataSourceMetadata<PartitionIdType, SequenceOffsetType> resetMetadata = (SeekableStreamDataSourceMetadata<PartitionIdType, SequenceOffsetType>) dataSourceMetadata;
        if (resetMetadata.getSeekableStreamSequenceNumbers().getStream().equals(ioConfig.getStream())) {
            // metadata can be null
            final DataSourceMetadata metadata = indexerMetadataStorageCoordinator.retrieveDataSourceMetadata(dataSource);
            if (metadata != null && !checkSourceMetadataMatch(metadata)) {
                throw new IAE("Datasource metadata instance does not match required, found instance of [%s]", metadata.getClass());
            }
            @SuppressWarnings("unchecked") final SeekableStreamDataSourceMetadata<PartitionIdType, SequenceOffsetType> currentMetadata = (SeekableStreamDataSourceMetadata<PartitionIdType, SequenceOffsetType>) metadata;
            // defend against consecutive reset requests from replicas
            // as well as the case where the metadata store do not have an entry for the reset partitions
            boolean doReset = false;
            for (Entry<PartitionIdType, SequenceOffsetType> resetPartitionOffset : resetMetadata.getSeekableStreamSequenceNumbers().getPartitionSequenceNumberMap().entrySet()) {
                final SequenceOffsetType partitionOffsetInMetadataStore = currentMetadata == null ? null : currentMetadata.getSeekableStreamSequenceNumbers().getPartitionSequenceNumberMap().get(resetPartitionOffset.getKey());
                final TaskGroup partitionTaskGroup = activelyReadingTaskGroups.get(getTaskGroupIdForPartition(resetPartitionOffset.getKey()));
                final boolean isSameOffset = partitionTaskGroup != null && partitionTaskGroup.startingSequences.get(resetPartitionOffset.getKey()).equals(resetPartitionOffset.getValue());
                if (partitionOffsetInMetadataStore != null || isSameOffset) {
                    doReset = true;
                    break;
                }
            }
            if (!doReset) {
                log.info("Ignoring duplicate reset request [%s]", dataSourceMetadata);
                return;
            }
            boolean metadataUpdateSuccess;
            if (currentMetadata == null) {
                metadataUpdateSuccess = true;
            } else {
                final DataSourceMetadata newMetadata = currentMetadata.minus(resetMetadata);
                try {
                    metadataUpdateSuccess = indexerMetadataStorageCoordinator.resetDataSourceMetadata(dataSource, newMetadata);
                } catch (IOException e) {
                    log.error("Resetting DataSourceMetadata failed [%s]", e.getMessage());
                    throw new RuntimeException(e);
                }
            }
            if (metadataUpdateSuccess) {
                resetMetadata.getSeekableStreamSequenceNumbers().getPartitionSequenceNumberMap().keySet().forEach(partition -> {
                    final int groupId = getTaskGroupIdForPartition(partition);
                    killTaskGroupForPartitions(ImmutableSet.of(partition), "DataSourceMetadata is updated while reset");
                    activelyReadingTaskGroups.remove(groupId);
                    // killTaskGroupForPartitions() cleans up partitionGroups.
                    // Add the removed groups back.
                    partitionGroups.computeIfAbsent(groupId, k -> new HashSet<>());
                    partitionOffsets.put(partition, getNotSetMarker());
                });
            } else {
                throw new ISE("Unable to reset metadata");
            }
        } else {
            log.warn("Reset metadata stream [%s] and supervisor's stream name [%s] do not match", resetMetadata.getSeekableStreamSequenceNumbers().getStream(), ioConfig.getStream());
        }
    }
}

Example 3 with DataSourceMetadata

use of org.apache.druid.indexing.overlord.DataSourceMetadata in project druid by druid-io.

the class AppenderatorDriverRealtimeIndexTaskTest method makeToolboxFactory.

private void makeToolboxFactory(final File directory) {
    taskStorage = new HeapMemoryTaskStorage(new TaskStorageConfig(null));
    publishedSegments = new CopyOnWriteArrayList<>();
    ObjectMapper mapper = new DefaultObjectMapper();
    mapper.registerSubtypes(LinearShardSpec.class);
    mapper.registerSubtypes(NumberedShardSpec.class);
    IndexerSQLMetadataStorageCoordinator mdc = new IndexerSQLMetadataStorageCoordinator(mapper, derbyConnectorRule.metadataTablesConfigSupplier().get(), derbyConnectorRule.getConnector()) {

        @Override
        public Set<DataSegment> announceHistoricalSegments(Set<DataSegment> segments) throws IOException {
            Set<DataSegment> result = super.announceHistoricalSegments(segments);
            Assert.assertFalse("Segment latch not initialized, did you forget to call expectPublishSegments?", segmentLatch == null);
            publishedSegments.addAll(result);
            segments.forEach(s -> segmentLatch.countDown());
            return result;
        }

        @Override
        public SegmentPublishResult announceHistoricalSegments(Set<DataSegment> segments, Set<DataSegment> segmentsToDrop, DataSourceMetadata startMetadata, DataSourceMetadata endMetadata) throws IOException {
            SegmentPublishResult result = super.announceHistoricalSegments(segments, segmentsToDrop, startMetadata, endMetadata);
            Assert.assertFalse("Segment latch not initialized, did you forget to call expectPublishSegments?", segmentLatch == null);
            publishedSegments.addAll(result.getSegments());
            result.getSegments().forEach(s -> segmentLatch.countDown());
            return result;
        }
    };
    taskLockbox = new TaskLockbox(taskStorage, mdc);
    final TaskConfig taskConfig = new TaskConfig(directory.getPath(), null, null, 50000, null, true, null, null, null, false, false, TaskConfig.BATCH_PROCESSING_MODE_DEFAULT.name());
    final TaskActionToolbox taskActionToolbox = new TaskActionToolbox(taskLockbox, taskStorage, mdc, EMITTER, EasyMock.createMock(SupervisorManager.class));
    final TaskActionClientFactory taskActionClientFactory = new LocalTaskActionClientFactory(taskStorage, taskActionToolbox, new TaskAuditLogConfig(false));
    final QueryRunnerFactoryConglomerate conglomerate = new DefaultQueryRunnerFactoryConglomerate(ImmutableMap.of(TimeseriesQuery.class, new TimeseriesQueryRunnerFactory(new TimeseriesQueryQueryToolChest(), new TimeseriesQueryEngine(), (query, future) -> {
    // do nothing
    })));
    handOffCallbacks = new ConcurrentHashMap<>();
    final SegmentHandoffNotifierFactory handoffNotifierFactory = dataSource -> new SegmentHandoffNotifier() {

        @Override
        public boolean registerSegmentHandoffCallback(SegmentDescriptor descriptor, Executor exec, Runnable handOffRunnable) {
            handOffCallbacks.put(descriptor, new Pair<>(exec, handOffRunnable));
            handoffLatch.countDown();
            return true;
        }

        @Override
        public void start() {
        // Noop
        }

        @Override
        public void close() {
        // Noop
        }
    };
    final TestUtils testUtils = new TestUtils();
    taskToolboxFactory = new TaskToolboxFactory(taskConfig, new DruidNode("druid/middlemanager", "localhost", false, 8091, null, true, false), taskActionClientFactory, EMITTER, new TestDataSegmentPusher(), new TestDataSegmentKiller(), // DataSegmentMover
    null, // DataSegmentArchiver
    null, new TestDataSegmentAnnouncer(), EasyMock.createNiceMock(DataSegmentServerAnnouncer.class), handoffNotifierFactory, () -> conglomerate, // queryExecutorService
    DirectQueryProcessingPool.INSTANCE, NoopJoinableFactory.INSTANCE, () -> EasyMock.createMock(MonitorScheduler.class), new SegmentCacheManagerFactory(testUtils.getTestObjectMapper()), testUtils.getTestObjectMapper(), testUtils.getTestIndexIO(), MapCache.create(1024), new CacheConfig(), new CachePopulatorStats(), testUtils.getTestIndexMergerV9(), EasyMock.createNiceMock(DruidNodeAnnouncer.class), EasyMock.createNiceMock(DruidNode.class), new LookupNodeService("tier"), new DataNodeService("tier", 1000, ServerType.INDEXER_EXECUTOR, 0), new SingleFileTaskReportFileWriter(reportsFile), null, AuthTestUtils.TEST_AUTHORIZER_MAPPER, new NoopChatHandlerProvider(), testUtils.getRowIngestionMetersFactory(), new TestAppenderatorsManager(), new NoopIndexingServiceClient(), null, null, null);
}

Example 4 with DataSourceMetadata

use of org.apache.druid.indexing.overlord.DataSourceMetadata in project druid by druid-io.

the class DerivativeDataSourceManager method updateDerivatives.

private void updateDerivatives() {
    List<Pair<String, DerivativeDataSourceMetadata>> derivativesInDatabase = connector.retryWithHandle(handle -> handle.createQuery(StringUtils.format("SELECT DISTINCT dataSource,commit_metadata_payload FROM %1$s", dbTables.get().getDataSourceTable())).map((int index, ResultSet r, StatementContext ctx) -> {
        String datasourceName = r.getString("dataSource");
        DataSourceMetadata payload = JacksonUtils.readValue(objectMapper, r.getBytes("commit_metadata_payload"), DataSourceMetadata.class);
        if (!(payload instanceof DerivativeDataSourceMetadata)) {
            return null;
        }
        DerivativeDataSourceMetadata metadata = (DerivativeDataSourceMetadata) payload;
        return new Pair<>(datasourceName, metadata);
    }).list());
    List<DerivativeDataSource> derivativeDataSources = derivativesInDatabase.parallelStream().filter(data -> data != null).map(derivatives -> {
        String name = derivatives.lhs;
        DerivativeDataSourceMetadata metadata = derivatives.rhs;
        String baseDataSource = metadata.getBaseDataSource();
        long avgSizePerGranularity = getAvgSizePerGranularity(name);
        log.info("find derivatives: {bases=%s, derivative=%s, dimensions=%s, metrics=%s, avgSize=%s}", baseDataSource, name, metadata.getDimensions(), metadata.getMetrics(), avgSizePerGranularity);
        return new DerivativeDataSource(name, baseDataSource, metadata.getColumns(), avgSizePerGranularity);
    }).filter(derivatives -> derivatives.getAvgSizeBasedGranularity() > 0).collect(Collectors.toList());
    ConcurrentHashMap<String, SortedSet<DerivativeDataSource>> newDerivatives = new ConcurrentHashMap<>();
    for (DerivativeDataSource derivative : derivativeDataSources) {
        newDerivatives.computeIfAbsent(derivative.getBaseDataSource(), ds -> new TreeSet<>()).add(derivative);
    }
    ConcurrentHashMap<String, SortedSet<DerivativeDataSource>> current;
    do {
        current = DERIVATIVES_REF.get();
    } while (!DERIVATIVES_REF.compareAndSet(current, newDerivatives));
}

Example 5 with DataSourceMetadata

use of org.apache.druid.indexing.overlord.DataSourceMetadata in project druid by druid-io.

the class KinesisIndexTaskTest method makeToolboxFactory.

private void makeToolboxFactory() throws IOException {
    directory = tempFolder.newFolder();
    final TestUtils testUtils = new TestUtils();
    final ObjectMapper objectMapper = testUtils.getTestObjectMapper();
    objectMapper.setInjectableValues(((InjectableValues.Std) objectMapper.getInjectableValues()).addValue(AWSCredentialsConfig.class, new AWSCredentialsConfig()));
    for (Module module : new KinesisIndexingServiceModule().getJacksonModules()) {
        objectMapper.registerModule(module);
    }
    final TaskConfig taskConfig = new TaskConfig(new File(directory, "baseDir").getPath(), new File(directory, "baseTaskDir").getPath(), null, 50000, null, true, null, null, null, false, false, TaskConfig.BATCH_PROCESSING_MODE_DEFAULT.name());
    final TestDerbyConnector derbyConnector = derby.getConnector();
    derbyConnector.createDataSourceTable();
    derbyConnector.createPendingSegmentsTable();
    derbyConnector.createSegmentTable();
    derbyConnector.createRulesTable();
    derbyConnector.createConfigTable();
    derbyConnector.createTaskTables();
    derbyConnector.createAuditTable();
    taskStorage = new MetadataTaskStorage(derbyConnector, new TaskStorageConfig(null), new DerbyMetadataStorageActionHandlerFactory(derbyConnector, derby.metadataTablesConfigSupplier().get(), objectMapper));
    metadataStorageCoordinator = new IndexerSQLMetadataStorageCoordinator(testUtils.getTestObjectMapper(), derby.metadataTablesConfigSupplier().get(), derbyConnector);
    taskLockbox = new TaskLockbox(taskStorage, metadataStorageCoordinator);
    final TaskActionToolbox taskActionToolbox = new TaskActionToolbox(taskLockbox, taskStorage, metadataStorageCoordinator, emitter, new SupervisorManager(null) {

        @Override
        public boolean checkPointDataSourceMetadata(String supervisorId, int taskGroupId, @Nullable DataSourceMetadata checkpointMetadata) {
            LOG.info("Adding checkpoint hash to the set");
            checkpointRequestsHash.add(Objects.hash(supervisorId, taskGroupId, checkpointMetadata));
            return true;
        }
    });
    final TaskActionClientFactory taskActionClientFactory = new LocalTaskActionClientFactory(taskStorage, taskActionToolbox, new TaskAuditLogConfig(false));
    final SegmentHandoffNotifierFactory handoffNotifierFactory = dataSource -> new SegmentHandoffNotifier() {

        @Override
        public boolean registerSegmentHandoffCallback(SegmentDescriptor descriptor, Executor exec, Runnable handOffRunnable) {
            if (doHandoff) {
                // Simulate immediate handoff
                exec.execute(handOffRunnable);
            }
            return true;
        }

        @Override
        public void start() {
        // Noop
        }

        @Override
        public void close() {
        // Noop
        }
    };
    final LocalDataSegmentPusherConfig dataSegmentPusherConfig = new LocalDataSegmentPusherConfig();
    dataSegmentPusherConfig.storageDirectory = getSegmentDirectory();
    final DataSegmentPusher dataSegmentPusher = new LocalDataSegmentPusher(dataSegmentPusherConfig);
    toolboxFactory = new TaskToolboxFactory(taskConfig, // taskExecutorNode
    null, taskActionClientFactory, emitter, dataSegmentPusher, new TestDataSegmentKiller(), // DataSegmentMover
    null, // DataSegmentArchiver
    null, new TestDataSegmentAnnouncer(), EasyMock.createNiceMock(DataSegmentServerAnnouncer.class), handoffNotifierFactory, this::makeTimeseriesOnlyConglomerate, DirectQueryProcessingPool.INSTANCE, NoopJoinableFactory.INSTANCE, () -> EasyMock.createMock(MonitorScheduler.class), new SegmentCacheManagerFactory(testUtils.getTestObjectMapper()), testUtils.getTestObjectMapper(), testUtils.getTestIndexIO(), MapCache.create(1024), new CacheConfig(), new CachePopulatorStats(), testUtils.getTestIndexMergerV9(), EasyMock.createNiceMock(DruidNodeAnnouncer.class), EasyMock.createNiceMock(DruidNode.class), new LookupNodeService("tier"), new DataNodeService("tier", 1, ServerType.INDEXER_EXECUTOR, 0), new SingleFileTaskReportFileWriter(reportsFile), null, AuthTestUtils.TEST_AUTHORIZER_MAPPER, new NoopChatHandlerProvider(), testUtils.getRowIngestionMetersFactory(), new TestAppenderatorsManager(), new NoopIndexingServiceClient(), null, null, null);
}