use of org.apache.hyracks.algebricks.common.exceptions.AlgebricksException in project asterixdb by apache.
the class InvertedIndexPOperator method buildInvertedIndexRuntime.
public Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> buildInvertedIndexRuntime(MetadataProvider metadataProvider, JobGenContext context, JobSpecification jobSpec, AbstractUnnestMapOperator unnestMap, IOperatorSchema opSchema, boolean retainInput, boolean retainMissing, String datasetName, Dataset dataset, String indexName, ATypeTag searchKeyType, int[] keyFields, SearchModifierType searchModifierType, IAlgebricksConstantValue similarityThreshold, int[] minFilterFieldIndexes, int[] maxFilterFieldIndexes, boolean isFullTextSearchQuery) throws AlgebricksException {
try {
IAObject simThresh = ((AsterixConstantValue) similarityThreshold).getObject();
int numPrimaryKeys = dataset.getPrimaryKeys().size();
Index secondaryIndex = MetadataManager.INSTANCE.getIndex(metadataProvider.getMetadataTxnContext(), dataset.getDataverseName(), dataset.getDatasetName(), indexName);
if (secondaryIndex == null) {
throw new AlgebricksException("Code generation error: no index " + indexName + " for dataset " + datasetName);
}
IVariableTypeEnvironment typeEnv = context.getTypeEnvironment(unnestMap);
RecordDescriptor outputRecDesc = JobGenHelper.mkRecordDescriptor(typeEnv, opSchema, context);
Pair<IFileSplitProvider, AlgebricksPartitionConstraint> secondarySplitsAndConstraint = metadataProvider.getSplitProviderAndConstraints(dataset, indexName);
// TODO: Here we assume there is only one search key field.
int queryField = keyFields[0];
// Get tokenizer and search modifier factories.
IInvertedIndexSearchModifierFactory searchModifierFactory = InvertedIndexAccessMethod.getSearchModifierFactory(searchModifierType, simThresh, secondaryIndex);
IBinaryTokenizerFactory queryTokenizerFactory = InvertedIndexAccessMethod.getBinaryTokenizerFactory(searchModifierType, searchKeyType, secondaryIndex);
IIndexDataflowHelperFactory dataflowHelperFactory = new IndexDataflowHelperFactory(metadataProvider.getStorageComponentProvider().getStorageManager(), secondarySplitsAndConstraint.first);
LSMInvertedIndexSearchOperatorDescriptor invIndexSearchOp = new LSMInvertedIndexSearchOperatorDescriptor(jobSpec, outputRecDesc, queryField, dataflowHelperFactory, queryTokenizerFactory, searchModifierFactory, retainInput, retainMissing, context.getMissingWriterFactory(), dataset.getSearchCallbackFactory(metadataProvider.getStorageComponentProvider(), secondaryIndex, ((JobEventListenerFactory) jobSpec.getJobletEventListenerFactory()).getJobId(), IndexOperation.SEARCH, null), minFilterFieldIndexes, maxFilterFieldIndexes, isFullTextSearchQuery, numPrimaryKeys, false);
return new Pair<>(invIndexSearchOp, secondarySplitsAndConstraint.second);
} catch (MetadataException e) {
throw new AlgebricksException(e);
}
}
use of org.apache.hyracks.algebricks.common.exceptions.AlgebricksException in project asterixdb by apache.
the class QueryLogicalExpressionJobGen method createSerializableAggregateFunctionFactory.
@Override
public ISerializedAggregateEvaluatorFactory createSerializableAggregateFunctionFactory(AggregateFunctionCallExpression expr, IVariableTypeEnvironment env, IOperatorSchema[] inputSchemas, JobGenContext context) throws AlgebricksException {
IScalarEvaluatorFactory[] args = codegenArguments(expr, env, inputSchemas, context);
IFunctionDescriptor fd = getFunctionDescriptor(expr, env, context);
switch(fd.getFunctionDescriptorTag()) {
case AGGREGATE:
{
if (BuiltinFunctions.isAggregateFunctionSerializable(fd.getIdentifier())) {
AggregateFunctionCallExpression serialAggExpr = BuiltinFunctions.makeSerializableAggregateFunctionExpression(fd.getIdentifier(), expr.getArguments());
IFunctionDescriptor afdd = getFunctionDescriptor(serialAggExpr, env, context);
return afdd.createSerializableAggregateEvaluatorFactory(args);
} else {
throw new AlgebricksException("Trying to create a serializable aggregate from a non-serializable aggregate function descriptor. (fi=" + expr.getFunctionIdentifier() + ")");
}
}
case SERIALAGGREGATE:
{
return fd.createSerializableAggregateEvaluatorFactory(args);
}
default:
throw new IllegalStateException("Invalid function descriptor " + fd.getFunctionDescriptorTag() + " expected " + FunctionDescriptorTag.SERIALAGGREGATE + " or " + FunctionDescriptorTag.AGGREGATE);
}
}
use of org.apache.hyracks.algebricks.common.exceptions.AlgebricksException in project asterixdb by apache.
the class QueryTranslator method handleNodegroupDropStatement.
protected void handleNodegroupDropStatement(MetadataProvider metadataProvider, Statement stmt) throws Exception {
NodeGroupDropStatement stmtDelete = (NodeGroupDropStatement) stmt;
String nodegroupName = stmtDelete.getNodeGroupName().getValue();
MetadataTransactionContext mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
metadataProvider.setMetadataTxnContext(mdTxnCtx);
MetadataLockManager.INSTANCE.acquireNodeGroupWriteLock(metadataProvider.getLocks(), nodegroupName);
try {
NodeGroup ng = MetadataManager.INSTANCE.getNodegroup(mdTxnCtx, nodegroupName);
if (ng == null) {
if (!stmtDelete.getIfExists()) {
throw new AlgebricksException("There is no nodegroup with this name " + nodegroupName + ".");
}
} else {
MetadataManager.INSTANCE.dropNodegroup(mdTxnCtx, nodegroupName, false);
}
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
} catch (Exception e) {
abort(e, e, mdTxnCtx);
throw e;
} finally {
metadataProvider.getLocks().unlock();
}
}
use of org.apache.hyracks.algebricks.common.exceptions.AlgebricksException in project asterixdb by apache.
the class QueryTranslator method handleExternalDatasetRefreshStatement.
protected void handleExternalDatasetRefreshStatement(MetadataProvider metadataProvider, Statement stmt, IHyracksClientConnection hcc) throws Exception {
RefreshExternalDatasetStatement stmtRefresh = (RefreshExternalDatasetStatement) stmt;
String dataverseName = getActiveDataverse(stmtRefresh.getDataverseName());
String datasetName = stmtRefresh.getDatasetName().getValue();
TransactionState transactionState = TransactionState.COMMIT;
MetadataTransactionContext mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
boolean bActiveTxn = true;
metadataProvider.setMetadataTxnContext(mdTxnCtx);
JobSpecification spec = null;
Dataset ds = null;
List<ExternalFile> metadataFiles = null;
List<ExternalFile> deletedFiles = null;
List<ExternalFile> addedFiles = null;
List<ExternalFile> appendedFiles = null;
List<Index> indexes = null;
Dataset transactionDataset = null;
boolean lockAquired = false;
boolean success = false;
MetadataLockManager.INSTANCE.refreshDatasetBegin(metadataProvider.getLocks(), dataverseName, dataverseName + "." + datasetName);
try {
ds = metadataProvider.findDataset(dataverseName, datasetName);
// Dataset exists ?
if (ds == null) {
throw new AlgebricksException("There is no dataset with this name " + datasetName + " in dataverse " + dataverseName);
}
// Dataset external ?
if (ds.getDatasetType() != DatasetType.EXTERNAL) {
throw new AlgebricksException("dataset " + datasetName + " in dataverse " + dataverseName + " is not an external dataset");
}
// Dataset has indexes ?
indexes = MetadataManager.INSTANCE.getDatasetIndexes(mdTxnCtx, dataverseName, datasetName);
if (indexes.isEmpty()) {
throw new AlgebricksException("External dataset " + datasetName + " in dataverse " + dataverseName + " doesn't have any index");
}
// Record transaction time
Date txnTime = new Date();
// refresh lock here
ExternalDatasetsRegistry.INSTANCE.refreshBegin(ds);
lockAquired = true;
// Get internal files
metadataFiles = MetadataManager.INSTANCE.getDatasetExternalFiles(mdTxnCtx, ds);
deletedFiles = new ArrayList<>();
addedFiles = new ArrayList<>();
appendedFiles = new ArrayList<>();
// Now we compare snapshot with external file system
if (ExternalIndexingOperations.isDatasetUptodate(ds, metadataFiles, addedFiles, deletedFiles, appendedFiles)) {
((ExternalDatasetDetails) ds.getDatasetDetails()).setRefreshTimestamp(txnTime);
MetadataManager.INSTANCE.updateDataset(mdTxnCtx, ds);
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
// latch will be released in the finally clause
return;
}
// At this point, we know data has changed in the external file system, record
// transaction in metadata and start
transactionDataset = ExternalIndexingOperations.createTransactionDataset(ds);
/*
* Remove old dataset record and replace it with a new one
*/
MetadataManager.INSTANCE.updateDataset(mdTxnCtx, transactionDataset);
// Add delta files to the metadata
for (ExternalFile file : addedFiles) {
MetadataManager.INSTANCE.addExternalFile(mdTxnCtx, file);
}
for (ExternalFile file : appendedFiles) {
MetadataManager.INSTANCE.addExternalFile(mdTxnCtx, file);
}
for (ExternalFile file : deletedFiles) {
MetadataManager.INSTANCE.addExternalFile(mdTxnCtx, file);
}
// Create the files index update job
spec = ExternalIndexingOperations.buildFilesIndexUpdateOp(ds, metadataFiles, addedFiles, appendedFiles, metadataProvider);
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
bActiveTxn = false;
transactionState = TransactionState.BEGIN;
// run the files update job
JobUtils.runJob(hcc, spec, true);
for (Index index : indexes) {
if (!ExternalIndexingOperations.isFileIndex(index)) {
spec = ExternalIndexingOperations.buildIndexUpdateOp(ds, index, metadataFiles, addedFiles, appendedFiles, metadataProvider);
// run the files update job
JobUtils.runJob(hcc, spec, true);
}
}
// all index updates has completed successfully, record transaction state
spec = ExternalIndexingOperations.buildCommitJob(ds, indexes, metadataProvider);
// Aquire write latch again -> start a transaction and record the decision to commit
mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
metadataProvider.setMetadataTxnContext(mdTxnCtx);
bActiveTxn = true;
((ExternalDatasetDetails) transactionDataset.getDatasetDetails()).setState(TransactionState.READY_TO_COMMIT);
((ExternalDatasetDetails) transactionDataset.getDatasetDetails()).setRefreshTimestamp(txnTime);
MetadataManager.INSTANCE.updateDataset(mdTxnCtx, transactionDataset);
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
bActiveTxn = false;
transactionState = TransactionState.READY_TO_COMMIT;
// We don't release the latch since this job is expected to be quick
JobUtils.runJob(hcc, spec, true);
// Start a new metadata transaction to record the final state of the transaction
mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
metadataProvider.setMetadataTxnContext(mdTxnCtx);
bActiveTxn = true;
for (ExternalFile file : metadataFiles) {
if (file.getPendingOp() == ExternalFilePendingOp.DROP_OP) {
MetadataManager.INSTANCE.dropExternalFile(mdTxnCtx, file);
} else if (file.getPendingOp() == ExternalFilePendingOp.NO_OP) {
Iterator<ExternalFile> iterator = appendedFiles.iterator();
while (iterator.hasNext()) {
ExternalFile appendedFile = iterator.next();
if (file.getFileName().equals(appendedFile.getFileName())) {
// delete existing file
MetadataManager.INSTANCE.dropExternalFile(mdTxnCtx, file);
// delete existing appended file
MetadataManager.INSTANCE.dropExternalFile(mdTxnCtx, appendedFile);
// add the original file with appended information
appendedFile.setFileNumber(file.getFileNumber());
appendedFile.setPendingOp(ExternalFilePendingOp.NO_OP);
MetadataManager.INSTANCE.addExternalFile(mdTxnCtx, appendedFile);
iterator.remove();
}
}
}
}
// remove the deleted files delta
for (ExternalFile file : deletedFiles) {
MetadataManager.INSTANCE.dropExternalFile(mdTxnCtx, file);
}
// insert new files
for (ExternalFile file : addedFiles) {
MetadataManager.INSTANCE.dropExternalFile(mdTxnCtx, file);
file.setPendingOp(ExternalFilePendingOp.NO_OP);
MetadataManager.INSTANCE.addExternalFile(mdTxnCtx, file);
}
// mark the transaction as complete
((ExternalDatasetDetails) transactionDataset.getDatasetDetails()).setState(TransactionState.COMMIT);
MetadataManager.INSTANCE.updateDataset(mdTxnCtx, transactionDataset);
// commit metadata transaction
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
success = true;
} catch (Exception e) {
if (bActiveTxn) {
abort(e, e, mdTxnCtx);
}
if (transactionState == TransactionState.READY_TO_COMMIT) {
throw new IllegalStateException("System is inconsistent state: commit of (" + dataverseName + "." + datasetName + ") refresh couldn't carry out the commit phase", e);
}
if (transactionState == TransactionState.COMMIT) {
// Nothing to do , everything should be clean
throw e;
}
if (transactionState == TransactionState.BEGIN) {
// transaction failed, need to do the following
// clean NCs removing transaction components
mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
bActiveTxn = true;
metadataProvider.setMetadataTxnContext(mdTxnCtx);
spec = ExternalIndexingOperations.buildAbortOp(ds, indexes, metadataProvider);
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
bActiveTxn = false;
try {
JobUtils.runJob(hcc, spec, true);
} catch (Exception e2) {
// This should never happen -- fix throw illegal
e.addSuppressed(e2);
throw new IllegalStateException("System is in inconsistent state. Failed to abort refresh", e);
}
// return the state of the dataset to committed
try {
mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
for (ExternalFile file : deletedFiles) {
MetadataManager.INSTANCE.dropExternalFile(mdTxnCtx, file);
}
for (ExternalFile file : addedFiles) {
MetadataManager.INSTANCE.dropExternalFile(mdTxnCtx, file);
}
for (ExternalFile file : appendedFiles) {
MetadataManager.INSTANCE.dropExternalFile(mdTxnCtx, file);
}
MetadataManager.INSTANCE.updateDataset(mdTxnCtx, ds);
// commit metadata transaction
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
} catch (Exception e2) {
abort(e, e2, mdTxnCtx);
e.addSuppressed(e2);
throw new IllegalStateException("System is in inconsistent state. Failed to drop delta files", e);
}
}
} finally {
if (lockAquired) {
ExternalDatasetsRegistry.INSTANCE.refreshEnd(ds, success);
}
metadataProvider.getLocks().unlock();
}
}
use of org.apache.hyracks.algebricks.common.exceptions.AlgebricksException in project asterixdb by apache.
the class QueryTranslator method handleIndexDropStatement.
protected void handleIndexDropStatement(MetadataProvider metadataProvider, Statement stmt, IHyracksClientConnection hcc) throws Exception {
IndexDropStatement stmtIndexDrop = (IndexDropStatement) stmt;
String datasetName = stmtIndexDrop.getDatasetName().getValue();
String dataverseName = getActiveDataverse(stmtIndexDrop.getDataverseName());
ProgressState progress = ProgressState.NO_PROGRESS;
MetadataTransactionContext mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
boolean bActiveTxn = true;
metadataProvider.setMetadataTxnContext(mdTxnCtx);
List<JobSpecification> jobsToExecute = new ArrayList<>();
MetadataLockManager.INSTANCE.dropIndexBegin(metadataProvider.getLocks(), dataverseName, dataverseName + "." + datasetName);
String indexName = null;
// For external index
boolean dropFilesIndex = false;
try {
Dataset ds = metadataProvider.findDataset(dataverseName, datasetName);
if (ds == null) {
throw new AlgebricksException("There is no dataset with this name " + datasetName + " in dataverse " + dataverseName);
}
ActiveLifecycleListener activeListener = (ActiveLifecycleListener) appCtx.getActiveLifecycleListener();
ActiveJobNotificationHandler activeEventHandler = activeListener.getNotificationHandler();
IActiveEntityEventsListener[] listeners = activeEventHandler.getEventListeners();
StringBuilder builder = null;
for (IActiveEntityEventsListener listener : listeners) {
if (listener.isEntityUsingDataset(ds)) {
if (builder == null) {
builder = new StringBuilder();
}
builder.append(new FeedConnectionId(listener.getEntityId(), datasetName) + "\n");
}
}
if (builder != null) {
throw new CompilationException("Dataset" + datasetName + " is currently being fed into by the following active entities: " + builder.toString());
}
if (ds.getDatasetType() == DatasetType.INTERNAL) {
indexName = stmtIndexDrop.getIndexName().getValue();
Index index = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataverseName, datasetName, indexName);
if (index == null) {
if (stmtIndexDrop.getIfExists()) {
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
return;
} else {
throw new AlgebricksException("There is no index with this name " + indexName + ".");
}
}
// #. prepare a job to drop the index in NC.
jobsToExecute.add(IndexUtil.buildDropIndexJobSpec(index, metadataProvider, ds));
// #. mark PendingDropOp on the existing index
MetadataManager.INSTANCE.dropIndex(mdTxnCtx, dataverseName, datasetName, indexName);
MetadataManager.INSTANCE.addIndex(mdTxnCtx, new Index(dataverseName, datasetName, indexName, index.getIndexType(), index.getKeyFieldNames(), index.getKeyFieldSourceIndicators(), index.getKeyFieldTypes(), index.isEnforcingKeyFileds(), index.isPrimaryIndex(), MetadataUtil.PENDING_DROP_OP));
// #. commit the existing transaction before calling runJob.
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
bActiveTxn = false;
progress = ProgressState.ADDED_PENDINGOP_RECORD_TO_METADATA;
for (JobSpecification jobSpec : jobsToExecute) {
JobUtils.runJob(hcc, jobSpec, true);
}
// #. begin a new transaction
mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
bActiveTxn = true;
metadataProvider.setMetadataTxnContext(mdTxnCtx);
// #. finally, delete the existing index
MetadataManager.INSTANCE.dropIndex(mdTxnCtx, dataverseName, datasetName, indexName);
} else {
// External dataset
indexName = stmtIndexDrop.getIndexName().getValue();
Index index = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataverseName, datasetName, indexName);
if (index == null) {
if (stmtIndexDrop.getIfExists()) {
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
return;
} else {
throw new AlgebricksException("There is no index with this name " + indexName + ".");
}
} else if (ExternalIndexingOperations.isFileIndex(index)) {
throw new AlgebricksException("Dropping a dataset's files index is not allowed.");
}
// #. prepare a job to drop the index in NC.
jobsToExecute.add(IndexUtil.buildDropIndexJobSpec(index, metadataProvider, ds));
List<Index> datasetIndexes = MetadataManager.INSTANCE.getDatasetIndexes(mdTxnCtx, dataverseName, datasetName);
if (datasetIndexes.size() == 2) {
dropFilesIndex = true;
// only one index + the files index, we need to delete both of the indexes
for (Index externalIndex : datasetIndexes) {
if (ExternalIndexingOperations.isFileIndex(externalIndex)) {
jobsToExecute.add(ExternalIndexingOperations.buildDropFilesIndexJobSpec(metadataProvider, ds));
// #. mark PendingDropOp on the existing files index
MetadataManager.INSTANCE.dropIndex(mdTxnCtx, dataverseName, datasetName, externalIndex.getIndexName());
MetadataManager.INSTANCE.addIndex(mdTxnCtx, new Index(dataverseName, datasetName, externalIndex.getIndexName(), externalIndex.getIndexType(), externalIndex.getKeyFieldNames(), externalIndex.getKeyFieldSourceIndicators(), index.getKeyFieldTypes(), index.isEnforcingKeyFileds(), externalIndex.isPrimaryIndex(), MetadataUtil.PENDING_DROP_OP));
}
}
}
// #. mark PendingDropOp on the existing index
MetadataManager.INSTANCE.dropIndex(mdTxnCtx, dataverseName, datasetName, indexName);
MetadataManager.INSTANCE.addIndex(mdTxnCtx, new Index(dataverseName, datasetName, indexName, index.getIndexType(), index.getKeyFieldNames(), index.getKeyFieldSourceIndicators(), index.getKeyFieldTypes(), index.isEnforcingKeyFileds(), index.isPrimaryIndex(), MetadataUtil.PENDING_DROP_OP));
// #. commit the existing transaction before calling runJob.
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
bActiveTxn = false;
progress = ProgressState.ADDED_PENDINGOP_RECORD_TO_METADATA;
for (JobSpecification jobSpec : jobsToExecute) {
JobUtils.runJob(hcc, jobSpec, true);
}
// #. begin a new transaction
mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
bActiveTxn = true;
metadataProvider.setMetadataTxnContext(mdTxnCtx);
// #. finally, delete the existing index
MetadataManager.INSTANCE.dropIndex(mdTxnCtx, dataverseName, datasetName, indexName);
if (dropFilesIndex) {
// delete the files index too
MetadataManager.INSTANCE.dropIndex(mdTxnCtx, dataverseName, datasetName, IndexingConstants.getFilesIndexName(datasetName));
MetadataManager.INSTANCE.dropDatasetExternalFiles(mdTxnCtx, ds);
ExternalDatasetsRegistry.INSTANCE.removeDatasetInfo(ds);
}
}
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
} catch (Exception e) {
if (bActiveTxn) {
abort(e, e, mdTxnCtx);
}
if (progress == ProgressState.ADDED_PENDINGOP_RECORD_TO_METADATA) {
// remove the all indexes in NC
try {
for (JobSpecification jobSpec : jobsToExecute) {
JobUtils.runJob(hcc, jobSpec, true);
}
} catch (Exception e2) {
// do no throw exception since still the metadata needs to be compensated.
e.addSuppressed(e2);
}
// remove the record from the metadata.
mdTxnCtx = MetadataManager.INSTANCE.beginTransaction();
metadataProvider.setMetadataTxnContext(mdTxnCtx);
try {
MetadataManager.INSTANCE.dropIndex(metadataProvider.getMetadataTxnContext(), dataverseName, datasetName, indexName);
if (dropFilesIndex) {
MetadataManager.INSTANCE.dropIndex(metadataProvider.getMetadataTxnContext(), dataverseName, datasetName, IndexingConstants.getFilesIndexName(datasetName));
}
MetadataManager.INSTANCE.commitTransaction(mdTxnCtx);
} catch (Exception e2) {
e.addSuppressed(e2);
abort(e, e2, mdTxnCtx);
throw new IllegalStateException("System is inconsistent state: pending index(" + dataverseName + "." + datasetName + "." + indexName + ") couldn't be removed from the metadata", e);
}
}
throw e;
} finally {
metadataProvider.getLocks().unlock();
ExternalDatasetsRegistry.INSTANCE.releaseAcquiredLocks(metadataProvider);
}
}
Aggregations