Search in sources :

Example 11 with InternalDatasetDetails

use of org.apache.asterix.metadata.entities.InternalDatasetDetails in project asterixdb by apache.

the class QueryTranslator method handleCreateDatasetStatement.

public void handleCreateDatasetStatement(MetadataProvider metadataProvider, Statement stmt, IHyracksClientConnection hcc) throws CompilationException, Exception {
    MutableObject<ProgressState> progress = new MutableObject<>(ProgressState.NO_PROGRESS);
    DatasetDecl dd = (DatasetDecl) stmt;
    String dataverseName = getActiveDataverse(dd.getDataverse());
    String datasetName = dd.getName().getValue();
    DatasetType dsType = dd.getDatasetType();
    String itemTypeDataverseName = getActiveDataverse(dd.getItemTypeDataverse());
    String itemTypeName = dd.getItemTypeName().getValue();
    String metaItemTypeDataverseName = getActiveDataverse(dd.getMetaItemTypeDataverse());
    String metaItemTypeName = dd.getMetaItemTypeName().getValue();
    Identifier ngNameId = dd.getNodegroupName();
    String nodegroupName = ngNameId == null ? null : ngNameId.getValue();
    String compactionPolicy = dd.getCompactionPolicy();
    Map<String, String> compactionPolicyProperties = dd.getCompactionPolicyProperties();
    boolean defaultCompactionPolicy = compactionPolicy == null;
    boolean temp = dd.getDatasetDetailsDecl().isTemp();
    MetadataTransactionContext mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
    boolean bActiveTxn = true;
    metadataProvider.setMetadataTxnContext(mdTxnCtx);
    MetadataLockManager.INSTANCE.createDatasetBegin(metadataProvider.getLocks(), dataverseName, itemTypeDataverseName, itemTypeDataverseName + "." + itemTypeName, metaItemTypeDataverseName, metaItemTypeDataverseName + "." + metaItemTypeName, nodegroupName, compactionPolicy, dataverseName + "." + datasetName, defaultCompactionPolicy);
    Dataset dataset = null;
    try {
        IDatasetDetails datasetDetails = null;
        Dataset ds = metadataProvider.findDataset(dataverseName, datasetName);
        if (ds != null) {
            if (dd.getIfNotExists()) {
                MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
                return;
            } else {
                throw new AlgebricksException("A dataset with this name " + datasetName + " already exists.");
            }
        }
        Datatype dt = MetadataManager.INSTANCE.getDatatype(metadataProvider.getMetadataTxnContext(), itemTypeDataverseName, itemTypeName);
        if (dt == null) {
            throw new AlgebricksException(": type " + itemTypeName + " could not be found.");
        }
        String ngName = ngNameId != null ? ngNameId.getValue() : configureNodegroupForDataset(appCtx, dd.getHints(), dataverseName, datasetName, metadataProvider);
        if (compactionPolicy == null) {
            compactionPolicy = GlobalConfig.DEFAULT_COMPACTION_POLICY_NAME;
            compactionPolicyProperties = GlobalConfig.DEFAULT_COMPACTION_POLICY_PROPERTIES;
        } else {
            validateCompactionPolicy(compactionPolicy, compactionPolicyProperties, mdTxnCtx, false);
        }
        switch(dd.getDatasetType()) {
            case INTERNAL:
                IAType itemType = dt.getDatatype();
                if (itemType.getTypeTag() != ATypeTag.OBJECT) {
                    throw new AlgebricksException("Dataset type has to be a record type.");
                }
                IAType metaItemType = null;
                if (metaItemTypeDataverseName != null && metaItemTypeName != null) {
                    metaItemType = metadataProvider.findType(metaItemTypeDataverseName, metaItemTypeName);
                }
                if (metaItemType != null && metaItemType.getTypeTag() != ATypeTag.OBJECT) {
                    throw new AlgebricksException("Dataset meta type has to be a record type.");
                }
                ARecordType metaRecType = (ARecordType) metaItemType;
                List<List<String>> partitioningExprs = ((InternalDetailsDecl) dd.getDatasetDetailsDecl()).getPartitioningExprs();
                List<Integer> keySourceIndicators = ((InternalDetailsDecl) dd.getDatasetDetailsDecl()).getKeySourceIndicators();
                boolean autogenerated = ((InternalDetailsDecl) dd.getDatasetDetailsDecl()).isAutogenerated();
                ARecordType aRecordType = (ARecordType) itemType;
                List<IAType> partitioningTypes = ValidateUtil.validatePartitioningExpressions(aRecordType, metaRecType, partitioningExprs, keySourceIndicators, autogenerated);
                List<String> filterField = ((InternalDetailsDecl) dd.getDatasetDetailsDecl()).getFilterField();
                if (filterField != null) {
                    ValidateUtil.validateFilterField(aRecordType, filterField);
                }
                if (compactionPolicy == null && filterField != null) {
                    // If the dataset has a filter and the user didn't specify a merge
                    // policy, then we will pick the
                    // correlated-prefix as the default merge policy.
                    compactionPolicy = GlobalConfig.DEFAULT_FILTERED_DATASET_COMPACTION_POLICY_NAME;
                    compactionPolicyProperties = GlobalConfig.DEFAULT_COMPACTION_POLICY_PROPERTIES;
                }
                datasetDetails = new InternalDatasetDetails(InternalDatasetDetails.FileStructure.BTREE, InternalDatasetDetails.PartitioningStrategy.HASH, partitioningExprs, partitioningExprs, keySourceIndicators, partitioningTypes, autogenerated, filterField, temp);
                break;
            case EXTERNAL:
                String adapter = ((ExternalDetailsDecl) dd.getDatasetDetailsDecl()).getAdapter();
                Map<String, String> properties = ((ExternalDetailsDecl) dd.getDatasetDetailsDecl()).getProperties();
                datasetDetails = new ExternalDatasetDetails(adapter, properties, new Date(), TransactionState.COMMIT);
                break;
            default:
                throw new CompilationException("Unknown datatype " + dd.getDatasetType());
        }
        // #. initialize DatasetIdFactory if it is not initialized.
        if (!DatasetIdFactory.isInitialized()) {
            DatasetIdFactory.initialize(MetadataManager.INSTANCE.getMostRecentDatasetId());
        }
        // #. add a new dataset with PendingAddOp
        dataset = new Dataset(dataverseName, datasetName, itemTypeDataverseName, itemTypeName, metaItemTypeDataverseName, metaItemTypeName, ngName, compactionPolicy, compactionPolicyProperties, datasetDetails, dd.getHints(), dsType, DatasetIdFactory.generateDatasetId(), MetadataUtil.PENDING_ADD_OP);
        MetadataManager.INSTANCE.addDataset(metadataProvider.getMetadataTxnContext(), dataset);
        if (dd.getDatasetType() == DatasetType.INTERNAL) {
            JobSpecification jobSpec = DatasetUtil.createDatasetJobSpec(dataset, metadataProvider);
            // #. make metadataTxn commit before calling runJob.
            MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
            bActiveTxn = false;
            progress.setValue(ProgressState.ADDED_PENDINGOP_RECORD_TO_METADATA);
            // #. runJob
            JobUtils.runJob(hcc, jobSpec, true);
            // #. begin new metadataTxn
            mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
            bActiveTxn = true;
            metadataProvider.setMetadataTxnContext(mdTxnCtx);
        }
        // #. add a new dataset with PendingNoOp after deleting the dataset with PendingAddOp
        MetadataManager.INSTANCE.dropDataset(metadataProvider.getMetadataTxnContext(), dataverseName, datasetName);
        dataset.setPendingOp(MetadataUtil.PENDING_NO_OP);
        MetadataManager.INSTANCE.addDataset(metadataProvider.getMetadataTxnContext(), dataset);
        MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
    } catch (Exception e) {
        if (bActiveTxn) {
            abort(e, e, mdTxnCtx);
        }
        if (progress.getValue() == ProgressState.ADDED_PENDINGOP_RECORD_TO_METADATA) {
            // #. execute compensation operations
            // remove the index in NC
            // [Notice]
            // As long as we updated(and committed) metadata, we should remove any effect of the job
            // because an exception occurs during runJob.
            mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
            bActiveTxn = true;
            metadataProvider.setMetadataTxnContext(mdTxnCtx);
            try {
                JobSpecification jobSpec = DatasetUtil.dropDatasetJobSpec(dataset, metadataProvider);
                MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
                bActiveTxn = false;
                JobUtils.runJob(hcc, jobSpec, true);
            } catch (Exception e2) {
                e.addSuppressed(e2);
                if (bActiveTxn) {
                    abort(e, e2, mdTxnCtx);
                }
            }
            // remove the record from the metadata.
            mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
            metadataProvider.setMetadataTxnContext(mdTxnCtx);
            try {
                MetadataManager.INSTANCE.dropDataset(metadataProvider.getMetadataTxnContext(), dataverseName, datasetName);
                MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
            } catch (Exception e2) {
                e.addSuppressed(e2);
                abort(e, e2, mdTxnCtx);
                throw new IllegalStateException("System is inconsistent state: pending dataset(" + dataverseName + "." + datasetName + ") couldn't be removed from the metadata", e);
            }
        }
        throw e;
    } finally {
        metadataProvider.getLocks().unlock();
    }
}
Also used : ProgressState(org.apache.asterix.common.utils.JobUtils.ProgressState) MetadataTransactionContext(org.apache.asterix.metadata.MetadataTransactionContext) ExternalDetailsDecl(org.apache.asterix.lang.common.statement.ExternalDetailsDecl) DatasetType(org.apache.asterix.common.config.DatasetConfig.DatasetType) IDatasetDetails(org.apache.asterix.metadata.IDatasetDetails) Datatype(org.apache.asterix.metadata.entities.Datatype) DatasetDecl(org.apache.asterix.lang.common.statement.DatasetDecl) Identifier(org.apache.asterix.lang.common.struct.Identifier) ExternalDatasetDetails(org.apache.asterix.metadata.entities.ExternalDatasetDetails) ArrayList(java.util.ArrayList) List(java.util.List) JobSpecification(org.apache.hyracks.api.job.JobSpecification) MutableObject(org.apache.commons.lang3.mutable.MutableObject) CompilationException(org.apache.asterix.common.exceptions.CompilationException) IHyracksDataset(org.apache.hyracks.api.dataset.IHyracksDataset) IDataset(org.apache.asterix.common.metadata.IDataset) Dataset(org.apache.asterix.metadata.entities.Dataset) InternalDatasetDetails(org.apache.asterix.metadata.entities.InternalDatasetDetails) AlgebricksException(org.apache.hyracks.algebricks.common.exceptions.AlgebricksException) Date(java.util.Date) ACIDException(org.apache.asterix.common.exceptions.ACIDException) MetadataException(org.apache.asterix.metadata.MetadataException) AlgebricksException(org.apache.hyracks.algebricks.common.exceptions.AlgebricksException) HyracksDataException(org.apache.hyracks.api.exceptions.HyracksDataException) CompilationException(org.apache.asterix.common.exceptions.CompilationException) IOException(java.io.IOException) RemoteException(java.rmi.RemoteException) AsterixException(org.apache.asterix.common.exceptions.AsterixException) InternalDetailsDecl(org.apache.asterix.lang.common.statement.InternalDetailsDecl) ARecordType(org.apache.asterix.om.types.ARecordType) IAType(org.apache.asterix.om.types.IAType)

Example 12 with InternalDatasetDetails

use of org.apache.asterix.metadata.entities.InternalDatasetDetails in project asterixdb by apache.

the class CheckpointingTest method testDeleteOldLogFiles.

@Test
public void testDeleteOldLogFiles() {
    try {
        TestNodeController nc = new TestNodeController(new File(TEST_CONFIG_FILE_PATH).getAbsolutePath(), false);
        StorageComponentProvider storageManager = new StorageComponentProvider();
        nc.init();
        List<List<String>> partitioningKeys = new ArrayList<>();
        partitioningKeys.add(Collections.singletonList("key"));
        Dataset dataset = new Dataset(DATAVERSE_NAME, DATASET_NAME, DATAVERSE_NAME, DATA_TYPE_NAME, NODE_GROUP_NAME, null, null, new InternalDatasetDetails(null, PartitioningStrategy.HASH, partitioningKeys, null, null, null, false, null, false), null, DatasetType.INTERNAL, DATASET_ID, 0);
        try {
            nc.createPrimaryIndex(dataset, KEY_TYPES, RECORD_TYPE, META_TYPE, new NoMergePolicyFactory(), null, null, storageManager, KEY_INDEXES, KEY_INDICATOR_LIST);
            IHyracksTaskContext ctx = nc.createTestContext(false);
            nc.newJobId();
            ITransactionContext txnCtx = nc.getTransactionManager().getTransactionContext(nc.getTxnJobId(), true);
            // Prepare insert operation
            LSMInsertDeleteOperatorNodePushable insertOp = nc.getInsertPipeline(ctx, dataset, KEY_TYPES, RECORD_TYPE, META_TYPE, new NoMergePolicyFactory(), null, null, KEY_INDEXES, KEY_INDICATOR_LIST, storageManager).getLeft();
            insertOp.open();
            TupleGenerator tupleGenerator = new TupleGenerator(RECORD_TYPE, META_TYPE, KEY_INDEXES, KEY_INDICATOR, RECORD_GEN_FUNCTION, UNIQUE_RECORD_FIELDS, META_GEN_FUNCTION, UNIQUE_META_FIELDS);
            VSizeFrame frame = new VSizeFrame(ctx);
            FrameTupleAppender tupleAppender = new FrameTupleAppender(frame);
            IRecoveryManager recoveryManager = nc.getTransactionSubsystem().getRecoveryManager();
            ICheckpointManager checkpointManager = nc.getTransactionSubsystem().getCheckpointManager();
            LogManager logManager = (LogManager) nc.getTransactionSubsystem().getLogManager();
            // Number of log files after node startup should be one
            int numberOfLogFiles = logManager.getLogFileIds().size();
            Assert.assertEquals(1, numberOfLogFiles);
            // Low-water mark LSN
            long lowWaterMarkLSN = recoveryManager.getMinFirstLSN();
            // Low-water mark log file id
            long initialLowWaterMarkFileId = logManager.getLogFileId(lowWaterMarkLSN);
            // Initial Low-water mark should be in the only available log file
            Assert.assertEquals(initialLowWaterMarkFileId, logManager.getLogFileIds().get(0).longValue());
            // Insert records until a new log file is created
            while (logManager.getLogFileIds().size() == 1) {
                ITupleReference tuple = tupleGenerator.next();
                DataflowUtils.addTupleToFrame(tupleAppender, tuple, insertOp);
            }
            // Check if the new low-water mark is still in the initial low-water mark log file
            lowWaterMarkLSN = recoveryManager.getMinFirstLSN();
            long currentLowWaterMarkLogFileId = logManager.getLogFileId(lowWaterMarkLSN);
            if (currentLowWaterMarkLogFileId == initialLowWaterMarkFileId) {
                /*
                     * Make sure checkpoint will not delete the initial log file since
                     * the low-water mark is still in it (i.e. it is still required for
                     * recovery)
                     */
                int numberOfLogFilesBeforeCheckpoint = logManager.getLogFileIds().size();
                checkpointManager.tryCheckpoint(logManager.getAppendLSN());
                int numberOfLogFilesAfterCheckpoint = logManager.getLogFileIds().size();
                Assert.assertEquals(numberOfLogFilesBeforeCheckpoint, numberOfLogFilesAfterCheckpoint);
                /*
                     * Insert records until the low-water mark is not in the initialLowWaterMarkFileId
                     * either because of the asynchronous flush caused by the previous checkpoint or a flush
                     * due to the dataset memory budget getting full.
                     */
                while (currentLowWaterMarkLogFileId == initialLowWaterMarkFileId) {
                    ITupleReference tuple = tupleGenerator.next();
                    DataflowUtils.addTupleToFrame(tupleAppender, tuple, insertOp);
                    lowWaterMarkLSN = recoveryManager.getMinFirstLSN();
                    currentLowWaterMarkLogFileId = logManager.getLogFileId(lowWaterMarkLSN);
                }
            }
            /*
                 * At this point, the low-water mark is not in the initialLowWaterMarkFileId, so
                 * a checkpoint should delete it.
                 */
            checkpointManager.tryCheckpoint(recoveryManager.getMinFirstLSN());
            // Validate initialLowWaterMarkFileId was deleted
            for (Long fileId : logManager.getLogFileIds()) {
                Assert.assertNotEquals(initialLowWaterMarkFileId, fileId.longValue());
            }
            if (tupleAppender.getTupleCount() > 0) {
                tupleAppender.write(insertOp, true);
            }
            insertOp.close();
            nc.getTransactionManager().completedTransaction(txnCtx, DatasetId.NULL, -1, true);
        } finally {
            nc.deInit();
        }
    } catch (Throwable e) {
        e.printStackTrace();
        Assert.fail(e.getMessage());
    }
}
Also used : LSMInsertDeleteOperatorNodePushable(org.apache.asterix.common.dataflow.LSMInsertDeleteOperatorNodePushable) ICheckpointManager(org.apache.asterix.common.transactions.ICheckpointManager) Dataset(org.apache.asterix.metadata.entities.Dataset) InternalDatasetDetails(org.apache.asterix.metadata.entities.InternalDatasetDetails) ITransactionContext(org.apache.asterix.common.transactions.ITransactionContext) ArrayList(java.util.ArrayList) TupleGenerator(org.apache.asterix.app.data.gen.TupleGenerator) StorageComponentProvider(org.apache.asterix.file.StorageComponentProvider) VSizeFrame(org.apache.hyracks.api.comm.VSizeFrame) NoMergePolicyFactory(org.apache.hyracks.storage.am.lsm.common.impls.NoMergePolicyFactory) IRecoveryManager(org.apache.asterix.common.transactions.IRecoveryManager) IHyracksTaskContext(org.apache.hyracks.api.context.IHyracksTaskContext) TestNodeController(org.apache.asterix.app.bootstrap.TestNodeController) FrameTupleAppender(org.apache.hyracks.dataflow.common.comm.io.FrameTupleAppender) ITupleReference(org.apache.hyracks.dataflow.common.data.accessors.ITupleReference) ArrayList(java.util.ArrayList) List(java.util.List) File(java.io.File) LogManager(org.apache.asterix.transaction.management.service.logging.LogManager) Test(org.junit.Test)

Example 13 with InternalDatasetDetails

use of org.apache.asterix.metadata.entities.InternalDatasetDetails in project asterixdb by apache.

the class DatasetTupleTranslatorTest method test.

@Test
public void test() throws MetadataException, IOException {
    Integer[] indicators = { 0, 1, null };
    for (Integer indicator : indicators) {
        Map<String, String> compactionPolicyProperties = new HashMap<>();
        compactionPolicyProperties.put("max-mergable-component-size", "1073741824");
        compactionPolicyProperties.put("max-tolerance-component-count", "3");
        InternalDatasetDetails details = new InternalDatasetDetails(FileStructure.BTREE, PartitioningStrategy.HASH, Collections.singletonList(Collections.singletonList("row_id")), Collections.singletonList(Collections.singletonList("row_id")), indicator == null ? null : Collections.singletonList(indicator), Collections.singletonList(BuiltinType.AINT64), false, Collections.emptyList(), false);
        Dataset dataset = new Dataset("test", "log", "foo", "LogType", "CB", "MetaType", "DEFAULT_NG_ALL_NODES", "prefix", compactionPolicyProperties, details, Collections.emptyMap(), DatasetType.INTERNAL, 115, 0);
        DatasetTupleTranslator dtTranslator = new DatasetTupleTranslator(true);
        ITupleReference tuple = dtTranslator.getTupleFromMetadataEntity(dataset);
        Dataset deserializedDataset = dtTranslator.getMetadataEntityFromTuple(tuple);
        Assert.assertEquals(dataset.getMetaItemTypeDataverseName(), deserializedDataset.getMetaItemTypeDataverseName());
        Assert.assertEquals(dataset.getMetaItemTypeName(), deserializedDataset.getMetaItemTypeName());
        if (indicator == null) {
            Assert.assertEquals(Collections.singletonList(new Integer(0)), ((InternalDatasetDetails) deserializedDataset.getDatasetDetails()).getKeySourceIndicator());
        } else {
            Assert.assertEquals(((InternalDatasetDetails) dataset.getDatasetDetails()).getKeySourceIndicator(), ((InternalDatasetDetails) deserializedDataset.getDatasetDetails()).getKeySourceIndicator());
        }
    }
}
Also used : HashMap(java.util.HashMap) Dataset(org.apache.asterix.metadata.entities.Dataset) InternalDatasetDetails(org.apache.asterix.metadata.entities.InternalDatasetDetails) ITupleReference(org.apache.hyracks.dataflow.common.data.accessors.ITupleReference) Test(org.junit.Test)

Example 14 with InternalDatasetDetails

use of org.apache.asterix.metadata.entities.InternalDatasetDetails in project asterixdb by apache.

the class MetadataBootstrap method insertMetadataDatasets.

/**
     * Inserts a metadata dataset to the physical dataset index
     * Should be performed on a bootstrap of a new universe
     *
     * @param mdTxnCtx
     * @param indexes
     * @throws MetadataException
     */
public static void insertMetadataDatasets(MetadataTransactionContext mdTxnCtx, IMetadataIndex[] indexes) throws MetadataException {
    for (int i = 0; i < indexes.length; i++) {
        IDatasetDetails id = new InternalDatasetDetails(FileStructure.BTREE, PartitioningStrategy.HASH, indexes[i].getPartitioningExpr(), indexes[i].getPartitioningExpr(), null, indexes[i].getPartitioningExprType(), false, null, false);
        MetadataManager.INSTANCE.addDataset(mdTxnCtx, new Dataset(indexes[i].getDataverseName(), indexes[i].getIndexedDatasetName(), indexes[i].getDataverseName(), indexes[i].getPayloadRecordType().getTypeName(), indexes[i].getNodeGroupName(), GlobalConfig.DEFAULT_COMPACTION_POLICY_NAME, GlobalConfig.DEFAULT_COMPACTION_POLICY_PROPERTIES, id, new HashMap<String, String>(), DatasetType.INTERNAL, indexes[i].getDatasetId().getId(), MetadataUtil.PENDING_NO_OP));
    }
    if (LOGGER.isLoggable(Level.INFO)) {
        LOGGER.info("Finished inserting initial datasets.");
    }
}
Also used : HashMap(java.util.HashMap) Dataset(org.apache.asterix.metadata.entities.Dataset) InternalDatasetDetails(org.apache.asterix.metadata.entities.InternalDatasetDetails) IDatasetDetails(org.apache.asterix.metadata.IDatasetDetails)

Aggregations

InternalDatasetDetails (org.apache.asterix.metadata.entities.InternalDatasetDetails)14 Dataset (org.apache.asterix.metadata.entities.Dataset)8 ArrayList (java.util.ArrayList)7 List (java.util.List)7 HashMap (java.util.HashMap)6 ITupleReference (org.apache.hyracks.dataflow.common.data.accessors.ITupleReference)5 AString (org.apache.asterix.om.base.AString)4 ARecordType (org.apache.asterix.om.types.ARecordType)4 IAType (org.apache.asterix.om.types.IAType)4 Test (org.junit.Test)4 IDatasetDetails (org.apache.asterix.metadata.IDatasetDetails)3 DatasetDataSource (org.apache.asterix.metadata.declared.DatasetDataSource)3 Index (org.apache.asterix.metadata.entities.Index)3 AInt32 (org.apache.asterix.om.base.AInt32)3 ILogicalExpression (org.apache.hyracks.algebricks.core.algebra.base.ILogicalExpression)3 LogicalVariable (org.apache.hyracks.algebricks.core.algebra.base.LogicalVariable)3 ScalarFunctionCallExpression (org.apache.hyracks.algebricks.core.algebra.expressions.ScalarFunctionCallExpression)3 VariableReferenceExpression (org.apache.hyracks.algebricks.core.algebra.expressions.VariableReferenceExpression)3 Date (java.util.Date)2 CommitOperator (org.apache.asterix.algebra.operators.CommitOperator)2