Examples with TableSnapshot org.apache.hadoop.hive.ql.io.AcidUtils.TableSnapshot used on opensource projects

Example 6 with TableSnapshot

use of org.apache.hadoop.hive.ql.io.AcidUtils.TableSnapshot in project hive by apache.

the class Hive method alterPartition.

/**
 * Updates the existing partition metadata with the new metadata.
 *
 * @param dbName
 *          name of the exiting table's database
 * @param tblName
 *          name of the existing table
 * @param newPart
 *          new partition
 * @param environmentContext
 *          environment context for the method
 * @param transactional
 *          indicates this call is for transaction stats
 * @throws InvalidOperationException
 *           if the changes in metadata is not acceptable
 * @throws HiveException
 */
public void alterPartition(String catName, String dbName, String tblName, Partition newPart, EnvironmentContext environmentContext, boolean transactional) throws InvalidOperationException, HiveException {
    try {
        if (catName == null) {
            catName = getDefaultCatalog(conf);
        }
        validatePartition(newPart);
        String location = newPart.getLocation();
        if (location != null) {
            location = Utilities.getQualifiedPath(conf, new Path(location));
            newPart.setLocation(location);
        }
        if (environmentContext == null) {
            environmentContext = new EnvironmentContext();
        }
        AcidUtils.TableSnapshot tableSnapshot = null;
        if (transactional) {
            tableSnapshot = AcidUtils.getTableSnapshot(conf, newPart.getTable(), true);
            if (tableSnapshot != null) {
                newPart.getTPartition().setWriteId(tableSnapshot.getWriteId());
            } else {
                LOG.warn("Cannot get a table snapshot for " + tblName);
            }
        }
        getSynchronizedMSC().alter_partition(catName, dbName, tblName, newPart.getTPartition(), environmentContext, tableSnapshot == null ? null : tableSnapshot.getValidWriteIdList());
    } catch (MetaException e) {
        throw new HiveException("Unable to alter partition. " + e.getMessage(), e);
    } catch (TException e) {
        throw new HiveException("Unable to alter partition. " + e.getMessage(), e);
    }
}

Example 7 with TableSnapshot

use of org.apache.hadoop.hive.ql.io.AcidUtils.TableSnapshot in project hive by apache.

the class Hive method createPartition.

/**
 * Creates a partition.
 *
 * @param tbl
 *          table for which partition needs to be created
 * @param partSpec
 *          partition keys and their values
 * @return created partition object
 * @throws HiveException
 *           if table doesn't exist or partition already exists
 */
@VisibleForTesting
public Partition createPartition(Table tbl, Map<String, String> partSpec) throws HiveException {
    try {
        org.apache.hadoop.hive.metastore.api.Partition part = Partition.createMetaPartitionObject(tbl, partSpec, null);
        AcidUtils.TableSnapshot tableSnapshot = AcidUtils.getTableSnapshot(conf, tbl);
        part.setWriteId(tableSnapshot != null ? tableSnapshot.getWriteId() : 0);
        return new Partition(tbl, getMSC().add_partition(part));
    } catch (Exception e) {
        LOG.error("Failed createPartition", e);
        throw new HiveException(e);
    }
}

Example 8 with TableSnapshot

use of org.apache.hadoop.hive.ql.io.AcidUtils.TableSnapshot in project hive by apache.

the class Hive method renamePartition.

/**
 * Rename a old partition to new partition
 *
 * @param tbl
 *          existing table
 * @param oldPartSpec
 *          spec of old partition
 * @param newPart
 *          new partition
 * @throws HiveException
 */
public void renamePartition(Table tbl, Map<String, String> oldPartSpec, Partition newPart, long replWriteId) throws HiveException {
    try {
        Map<String, String> newPartSpec = newPart.getSpec();
        if (oldPartSpec.keySet().size() != tbl.getPartCols().size() || newPartSpec.keySet().size() != tbl.getPartCols().size()) {
            throw new HiveException("Unable to rename partition to the same name: number of partition cols don't match. ");
        }
        if (!oldPartSpec.keySet().equals(newPartSpec.keySet())) {
            throw new HiveException("Unable to rename partition to the same name: old and new partition cols don't match. ");
        }
        List<String> pvals = new ArrayList<String>();
        for (FieldSchema field : tbl.getPartCols()) {
            String val = oldPartSpec.get(field.getName());
            if (val == null || val.length() == 0) {
                throw new HiveException("get partition: Value for key " + field.getName() + " is null or empty");
            } else if (val != null) {
                pvals.add(val);
            }
        }
        String validWriteIds = null;
        if (AcidUtils.isTransactionalTable(tbl)) {
            TableSnapshot tableSnapshot;
            if (replWriteId > 0) {
                // We need a valid writeId list for a transactional table modification. During
                // replication we do not have a valid writeId list which was used to modify the table
                // on the source. But we know for sure that the writeId associated with it was valid
                // then (otherwise modification would have failed on the source). So use a valid
                // transaction list with only that writeId.
                ValidWriteIdList writeIds = new ValidReaderWriteIdList(TableName.getDbTable(tbl.getDbName(), tbl.getTableName()), new long[0], new BitSet(), replWriteId);
                tableSnapshot = new TableSnapshot(replWriteId, writeIds.writeToString());
            } else {
                // Set table snapshot to api.Table to make it persistent.
                tableSnapshot = AcidUtils.getTableSnapshot(conf, tbl, true);
            }
            if (tableSnapshot != null) {
                newPart.getTPartition().setWriteId(tableSnapshot.getWriteId());
                validWriteIds = tableSnapshot.getValidWriteIdList();
            }
        }
        String catName = (tbl.getCatalogName() != null) ? tbl.getCatalogName() : getDefaultCatalog(conf);
        getMSC().renamePartition(catName, tbl.getDbName(), tbl.getTableName(), pvals, newPart.getTPartition(), validWriteIds);
    } catch (InvalidOperationException e) {
        throw new HiveException("Unable to rename partition. " + e.getMessage(), e);
    } catch (MetaException e) {
        throw new HiveException("Unable to rename partition. " + e.getMessage(), e);
    } catch (TException e) {
        throw new HiveException("Unable to rename partition. " + e.getMessage(), e);
    }
}

Example 9 with TableSnapshot

use of org.apache.hadoop.hive.ql.io.AcidUtils.TableSnapshot in project hive by apache.

the class Hive method loadDynamicPartitions.

/**
 * Given a source directory name of the load path, load all dynamically generated partitions
 * into the specified table and return a list of strings that represent the dynamic partition
 * paths.
 * @param tbd table descriptor
 * @param numLB number of buckets
 * @param isAcid true if this is an ACID operation
 * @param writeId writeId, can be 0 unless isAcid == true
 * @param stmtId statementId
 * @param resetStatistics if true, reset statistics. Do not reset statistics otherwise.
 * @param operation ACID operation type
 * @param partitionDetailsMap full dynamic partition specification
 * @return partition map details (PartitionSpec and Partition)
 * @throws HiveException
 */
public Map<Map<String, String>, Partition> loadDynamicPartitions(final LoadTableDesc tbd, final int numLB, final boolean isAcid, final long writeId, final int stmtId, final boolean resetStatistics, final AcidUtils.Operation operation, Map<Path, PartitionDetails> partitionDetailsMap) throws HiveException {
    PerfLogger perfLogger = SessionState.getPerfLogger();
    perfLogger.perfLogBegin("MoveTask", PerfLogger.LOAD_DYNAMIC_PARTITIONS);
    final Path loadPath = tbd.getSourcePath();
    final Table tbl = getTable(tbd.getTable().getTableName());
    final Map<String, String> partSpec = tbd.getPartitionSpec();
    final AtomicInteger partitionsLoaded = new AtomicInteger(0);
    final boolean inPlaceEligible = conf.getLong("fs.trash.interval", 0) <= 0 && InPlaceUpdate.canRenderInPlace(conf) && !SessionState.getConsole().getIsSilent();
    final PrintStream ps = (inPlaceEligible) ? SessionState.getConsole().getInfoStream() : null;
    final SessionState parentSession = SessionState.get();
    List<Callable<Partition>> tasks = Lists.newLinkedList();
    boolean fetchPartitionInfo = true;
    final boolean scanPartitionsByName = HiveConf.getBoolVar(conf, HIVE_LOAD_DYNAMIC_PARTITIONS_SCAN_SPECIFIC_PARTITIONS);
    // for every dynamic partition
    if (scanPartitionsByName && !tbd.isDirectInsert() && !AcidUtils.isTransactionalTable(tbl)) {
        // Fetch only relevant partitions from HMS for checking old partitions
        List<String> partitionNames = new LinkedList<>();
        for (PartitionDetails details : partitionDetailsMap.values()) {
            if (details.fullSpec != null && !details.fullSpec.isEmpty()) {
                partitionNames.add(Warehouse.makeDynamicPartNameNoTrailingSeperator(details.fullSpec));
            }
        }
        List<Partition> partitions = Hive.get().getPartitionsByNames(tbl, partitionNames);
        for (Partition partition : partitions) {
            LOG.debug("HMS partition spec: {}", partition.getSpec());
            partitionDetailsMap.entrySet().parallelStream().filter(entry -> entry.getValue().fullSpec.equals(partition.getSpec())).findAny().ifPresent(entry -> {
                entry.getValue().partition = partition;
                entry.getValue().hasOldPartition = true;
            });
        }
        // no need to fetch partition again in tasks since we have already fetched partitions
        // info in getPartitionsByNames()
        fetchPartitionInfo = false;
    }
    boolean isTxnTable = AcidUtils.isTransactionalTable(tbl);
    AcidUtils.TableSnapshot tableSnapshot = isTxnTable ? getTableSnapshot(tbl, writeId) : null;
    for (Entry<Path, PartitionDetails> entry : partitionDetailsMap.entrySet()) {
        boolean getPartitionFromHms = fetchPartitionInfo;
        tasks.add(() -> {
            PartitionDetails partitionDetails = entry.getValue();
            Map<String, String> fullPartSpec = partitionDetails.fullSpec;
            try {
                SessionState.setCurrentSessionState(parentSession);
                if (getPartitionFromHms) {
                    // didn't fetch partition info from HMS. Getting from HMS now.
                    Partition existing = getPartition(tbl, fullPartSpec, false);
                    if (existing != null) {
                        partitionDetails.partition = existing;
                        partitionDetails.hasOldPartition = true;
                    }
                }
                LOG.info("New loading path = " + entry.getKey() + " withPartSpec " + fullPartSpec);
                Partition oldPartition = partitionDetails.partition;
                List<FileStatus> newFiles = null;
                if (partitionDetails.newFiles != null) {
                    // If we already know the files from the direct insert manifest, use them
                    newFiles = partitionDetails.newFiles;
                } else if (conf.getBoolVar(ConfVars.FIRE_EVENTS_FOR_DML) && !tbl.isTemporary() && oldPartition == null) {
                    // Otherwise only collect them, if we are going to fire write notifications
                    newFiles = Collections.synchronizedList(new ArrayList<>());
                }
                // load the partition
                Partition partition = loadPartitionInternal(entry.getKey(), tbl, fullPartSpec, oldPartition, tbd.getLoadFileType(), true, false, numLB > 0, false, isAcid, resetStatistics, writeId, stmtId, tbd.isInsertOverwrite(), isTxnTable, newFiles, tbd.isDirectInsert());
                // metastore
                if (tableSnapshot != null) {
                    partition.getTPartition().setWriteId(tableSnapshot.getWriteId());
                }
                partitionDetails.tableSnapshot = tableSnapshot;
                if (oldPartition == null) {
                    partitionDetails.newFiles = newFiles;
                    partitionDetails.partition = partition;
                }
                if (inPlaceEligible) {
                    synchronized (ps) {
                        InPlaceUpdate.rePositionCursor(ps);
                        partitionsLoaded.incrementAndGet();
                        InPlaceUpdate.reprintLine(ps, "Loaded : " + partitionsLoaded.get() + "/" + partitionDetailsMap.size() + " partitions.");
                    }
                }
                return partition;
            } catch (Exception e) {
                LOG.error("Exception when loading partition with parameters " + " partPath=" + entry.getKey() + ", " + " table=" + tbl.getTableName() + ", " + " partSpec=" + fullPartSpec + ", " + " loadFileType=" + tbd.getLoadFileType().toString() + ", " + " listBucketingLevel=" + numLB + ", " + " isAcid=" + isAcid + ", " + " resetStatistics=" + resetStatistics, e);
                throw e;
            } finally {
                // get(conf).getMSC can be called in this task, Close the HMS connection right after use, do not wait for finalizer to close it.
                closeCurrent();
            }
        });
    }
    int poolSize = conf.getInt(ConfVars.HIVE_LOAD_DYNAMIC_PARTITIONS_THREAD_COUNT.varname, 1);
    ExecutorService executor = Executors.newFixedThreadPool(poolSize, new ThreadFactoryBuilder().setDaemon(true).setNameFormat("load-dynamic-partitionsToAdd-%d").build());
    List<Future<Partition>> futures = Lists.newLinkedList();
    Map<Map<String, String>, Partition> result = Maps.newLinkedHashMap();
    try {
        futures = executor.invokeAll(tasks);
        LOG.info("Number of partitionsToAdd to be added is " + futures.size());
        for (Future<Partition> future : futures) {
            Partition partition = future.get();
            result.put(partition.getSpec(), partition);
        }
        // add new partitions in batch
        addPartitionsToMetastore(partitionDetailsMap.entrySet().stream().filter(entry -> !entry.getValue().hasOldPartition).map(entry -> entry.getValue().partition).collect(Collectors.toList()), resetStatistics, tbl, partitionDetailsMap.entrySet().stream().filter(entry -> !entry.getValue().hasOldPartition).map(entry -> entry.getValue().tableSnapshot).collect(Collectors.toList()));
        // For acid table, add the acid_write event with file list at the time of load itself. But
        // it should be done after partition is created.
        List<WriteNotificationLogRequest> requestList = new ArrayList<>();
        int maxBatchSize = conf.getIntVar(HIVE_WRITE_NOTIFICATION_MAX_BATCH_SIZE);
        for (Entry<Path, PartitionDetails> entry : partitionDetailsMap.entrySet()) {
            PartitionDetails partitionDetails = entry.getValue();
            if (isTxnTable && partitionDetails.newFiles != null) {
                addWriteNotificationLog(tbl, partitionDetails.fullSpec, partitionDetails.newFiles, writeId, requestList);
                if (requestList != null && requestList.size() >= maxBatchSize) {
                    // If the first call returns that the HMS does not supports batching, avoid batching
                    // for later requests.
                    boolean batchSupported = addWriteNotificationLogInBatch(tbl, requestList);
                    if (batchSupported) {
                        requestList.clear();
                    } else {
                        requestList = null;
                    }
                }
            }
        }
        if (requestList != null && requestList.size() > 0) {
            addWriteNotificationLogInBatch(tbl, requestList);
        }
        setStatsPropAndAlterPartitions(resetStatistics, tbl, partitionDetailsMap.entrySet().stream().filter(entry -> entry.getValue().hasOldPartition).map(entry -> entry.getValue().partition).collect(Collectors.toList()), tableSnapshot);
    } catch (InterruptedException | ExecutionException e) {
        throw new HiveException("Exception when loading " + partitionDetailsMap.size() + " partitions" + " in table " + tbl.getTableName() + " with loadPath=" + loadPath, e);
    } catch (TException e) {
        LOG.error("Failed loadDynamicPartitions", e);
        throw new HiveException(e);
    } catch (Exception e) {
        StringBuffer logMsg = new StringBuffer();
        logMsg.append("Exception when loading partitionsToAdd with parameters ");
        logMsg.append("partPaths=");
        partitionDetailsMap.keySet().forEach(path -> logMsg.append(path + ", "));
        logMsg.append("table=" + tbl.getTableName() + ", ").append("partSpec=" + partSpec + ", ").append("loadFileType=" + tbd.getLoadFileType().toString() + ", ").append("listBucketingLevel=" + numLB + ", ").append("isAcid=" + isAcid + ", ").append("resetStatistics=" + resetStatistics);
        LOG.error(logMsg.toString(), e);
        throw e;
    } finally {
        LOG.debug("Cancelling " + futures.size() + " dynamic loading tasks");
        executor.shutdownNow();
    }
    if (HiveConf.getBoolVar(conf, ConfVars.HIVE_IN_TEST) && HiveConf.getBoolVar(conf, ConfVars.HIVETESTMODEFAILLOADDYNAMICPARTITION)) {
        throw new HiveException(HiveConf.ConfVars.HIVETESTMODEFAILLOADDYNAMICPARTITION.name() + "=true");
    }
    try {
        if (isTxnTable) {
            List<String> partNames = result.values().stream().map(Partition::getName).collect(Collectors.toList());
            getMSC().addDynamicPartitions(parentSession.getTxnMgr().getCurrentTxnId(), writeId, tbl.getDbName(), tbl.getTableName(), partNames, AcidUtils.toDataOperationType(operation));
        }
        LOG.info("Loaded " + result.size() + "partitionsToAdd");
        perfLogger.perfLogEnd("MoveTask", PerfLogger.LOAD_DYNAMIC_PARTITIONS);
        return result;
    } catch (TException te) {
        LOG.error("Failed loadDynamicPartitions", te);
        throw new HiveException("Exception updating metastore for acid table " + tbd.getTable().getTableName() + " with partitions " + result.values(), te);
    }
}

Example 10 with TableSnapshot

use of org.apache.hadoop.hive.ql.io.AcidUtils.TableSnapshot in project hive by apache.

the class Hive method setPartitionColumnStatistics.

public boolean setPartitionColumnStatistics(SetPartitionsStatsRequest request) throws HiveException {
    try {
        ColumnStatistics colStat = request.getColStats().get(0);
        ColumnStatisticsDesc statsDesc = colStat.getStatsDesc();
        // list obtained from the source. Just use it.
        if (request.getWriteId() <= 0 || request.getValidWriteIdList() == null) {
            Table tbl = getTable(statsDesc.getDbName(), statsDesc.getTableName());
            AcidUtils.TableSnapshot tableSnapshot = AcidUtils.getTableSnapshot(conf, tbl, true);
            request.setValidWriteIdList(tableSnapshot != null ? tableSnapshot.getValidWriteIdList() : null);
            request.setWriteId(tableSnapshot != null ? tableSnapshot.getWriteId() : 0);
        }
        return getMSC().setPartitionColumnStatistics(request);
    } catch (Exception e) {
        LOG.debug("Failed setPartitionColumnStatistics", e);
        throw new HiveException(e);
    }
}