Search in sources :

Example 1 with TableDescriptorBuilder

use of co.cask.cdap.data2.util.hbase.TableDescriptorBuilder in project cdap by caskdata.

the class ConfigurationTable method write.

/**
   * Writes the {@link CConfiguration} instance as a new row to the HBase table.  The {@link Type} given is used as
   * the row key (allowing multiple configurations to be stored).  After the new configuration is written, this will
   * delete any configurations written with an earlier timestamp (to prevent removed values from being visible).
   * @param cConf The CConfiguration instance to store
   * @throws IOException If an error occurs while writing the configuration
   */
public void write(Type type, CConfiguration cConf) throws IOException {
    // must create the table if it doesn't exist
    HTable table = null;
    try (HBaseDDLExecutor ddlExecutor = new HBaseDDLExecutorFactory(cConf, hbaseConf).get()) {
        HBaseTableUtil tableUtil = new HBaseTableUtilFactory(cConf).get();
        TableId tableId = tableUtil.createHTableId(NamespaceId.SYSTEM, TABLE_NAME);
        ColumnFamilyDescriptorBuilder cfdBuilder = HBaseTableUtil.getColumnFamilyDescriptorBuilder(Bytes.toString(FAMILY), hbaseConf);
        TableDescriptorBuilder tdBuilder = HBaseTableUtil.getTableDescriptorBuilder(tableId, cConf).addColumnFamily(cfdBuilder.build());
        ddlExecutor.createTableIfNotExists(tdBuilder.build(), null);
        long now = System.currentTimeMillis();
        long previous = now - 1;
        byte[] typeBytes = Bytes.toBytes(type.name());
        LOG.info("Writing new config row with key " + type);
        // populate the configuration data
        table = tableUtil.createHTable(hbaseConf, tableId);
        table.setAutoFlush(false);
        Put p = new Put(typeBytes);
        for (Map.Entry<String, String> e : cConf) {
            p.add(FAMILY, Bytes.toBytes(e.getKey()), now, Bytes.toBytes(e.getValue()));
        }
        table.put(p);
        LOG.info("Deleting any configuration from " + previous + " or before");
        Delete d = new Delete(typeBytes);
        d.deleteFamily(FAMILY, previous);
        table.delete(d);
    } finally {
        if (table != null) {
            try {
                table.close();
            } catch (IOException ioe) {
                LOG.error("Error closing HBaseAdmin: " + ioe.getMessage(), ioe);
            }
        }
    }
}
Also used : HBaseDDLExecutor(co.cask.cdap.spi.hbase.HBaseDDLExecutor) TableId(co.cask.cdap.data2.util.TableId) Delete(org.apache.hadoop.hbase.client.Delete) IOException(java.io.IOException) HTable(org.apache.hadoop.hbase.client.HTable) Put(org.apache.hadoop.hbase.client.Put) Map(java.util.Map)

Example 2 with TableDescriptorBuilder

use of co.cask.cdap.data2.util.hbase.TableDescriptorBuilder in project cdap by caskdata.

the class HBaseStreamFileConsumerFactory method create.

@Override
protected StreamConsumer create(TableId tableId, StreamConfig streamConfig, ConsumerConfig consumerConfig, StreamConsumerStateStore stateStore, StreamConsumerState beginConsumerState, FileReader<StreamEventOffset, Iterable<StreamFileOffset>> reader, @Nullable ReadFilter extraFilter) throws IOException {
    int splits = cConf.getInt(Constants.Stream.CONSUMER_TABLE_PRESPLITS);
    AbstractRowKeyDistributor distributor = new RowKeyDistributorByHashPrefix(new RowKeyDistributorByHashPrefix.OneByteSimpleHash(splits));
    byte[][] splitKeys = HBaseTableUtil.getSplitKeys(splits, splits, distributor);
    TableId hBaseTableId = tableUtil.createHTableId(new NamespaceId(tableId.getNamespace()), tableId.getTableName());
    TableDescriptorBuilder tdBuilder = HBaseTableUtil.getTableDescriptorBuilder(hBaseTableId, cConf);
    ColumnFamilyDescriptorBuilder cfdBuilder = HBaseTableUtil.getColumnFamilyDescriptorBuilder(Bytes.toString(QueueEntryRow.COLUMN_FAMILY), hConf);
    tdBuilder.addColumnFamily(cfdBuilder.build());
    tdBuilder.addProperty(QueueConstants.DISTRIBUTOR_BUCKETS, Integer.toString(splits));
    try (HBaseDDLExecutor ddlExecutor = ddlExecutorFactory.get()) {
        ddlExecutor.createTableIfNotExists(tdBuilder.build(), splitKeys);
    }
    HTable hTable = tableUtil.createHTable(hConf, hBaseTableId);
    hTable.setWriteBufferSize(Constants.Stream.HBASE_WRITE_BUFFER_SIZE);
    hTable.setAutoFlushTo(false);
    return new HBaseStreamFileConsumer(cConf, streamConfig, consumerConfig, tableUtil, hTable, reader, stateStore, beginConsumerState, extraFilter, createKeyDistributor(hTable.getTableDescriptor()));
}
Also used : TableId(co.cask.cdap.data2.util.TableId) HBaseDDLExecutor(co.cask.cdap.spi.hbase.HBaseDDLExecutor) RowKeyDistributorByHashPrefix(co.cask.cdap.hbase.wd.RowKeyDistributorByHashPrefix) AbstractRowKeyDistributor(co.cask.cdap.hbase.wd.AbstractRowKeyDistributor) ColumnFamilyDescriptorBuilder(co.cask.cdap.data2.util.hbase.ColumnFamilyDescriptorBuilder) TableDescriptorBuilder(co.cask.cdap.data2.util.hbase.TableDescriptorBuilder) NamespaceId(co.cask.cdap.proto.id.NamespaceId) HTable(org.apache.hadoop.hbase.client.HTable)

Example 3 with TableDescriptorBuilder

use of co.cask.cdap.data2.util.hbase.TableDescriptorBuilder in project cdap by caskdata.

the class HBaseStreamConsumerStateStoreFactory method create.

@Override
public synchronized StreamConsumerStateStore create(StreamConfig streamConfig) throws IOException {
    NamespaceId namespace = streamConfig.getStreamId().getParent();
    TableId streamStateStoreTableId = StreamUtils.getStateStoreTableId(namespace);
    TableId hbaseTableId = tableUtil.createHTableId(new NamespaceId(streamStateStoreTableId.getNamespace()), streamStateStoreTableId.getTableName());
    boolean tableExist;
    try (HBaseAdmin admin = new HBaseAdmin(hConf)) {
        tableExist = tableUtil.tableExists(admin, hbaseTableId);
    }
    if (!tableExist) {
        try (HBaseDDLExecutor ddlExecutor = ddlExecutorFactory.get()) {
            TableDescriptorBuilder tdBuilder = HBaseTableUtil.getTableDescriptorBuilder(hbaseTableId, cConf);
            ColumnFamilyDescriptorBuilder cfdBuilder = HBaseTableUtil.getColumnFamilyDescriptorBuilder(Bytes.toString(QueueEntryRow.COLUMN_FAMILY), hConf);
            tdBuilder.addColumnFamily(cfdBuilder.build());
            ddlExecutor.createTableIfNotExists(tdBuilder.build(), null);
        }
    }
    HTable hTable = tableUtil.createHTable(hConf, hbaseTableId);
    hTable.setWriteBufferSize(Constants.Stream.HBASE_WRITE_BUFFER_SIZE);
    hTable.setAutoFlushTo(false);
    return new HBaseStreamConsumerStateStore(streamConfig, hTable);
}
Also used : TableId(co.cask.cdap.data2.util.TableId) HBaseDDLExecutor(co.cask.cdap.spi.hbase.HBaseDDLExecutor) HBaseAdmin(org.apache.hadoop.hbase.client.HBaseAdmin) ColumnFamilyDescriptorBuilder(co.cask.cdap.data2.util.hbase.ColumnFamilyDescriptorBuilder) TableDescriptorBuilder(co.cask.cdap.data2.util.hbase.TableDescriptorBuilder) NamespaceId(co.cask.cdap.proto.id.NamespaceId) HTable(org.apache.hadoop.hbase.client.HTable)

Example 4 with TableDescriptorBuilder

use of co.cask.cdap.data2.util.hbase.TableDescriptorBuilder in project cdap by caskdata.

the class HBaseTableFactory method createMetadataTable.

@Override
public MetadataTable createMetadataTable(String tableName) throws IOException {
    TableId tableId = tableUtil.createHTableId(NamespaceId.SYSTEM, tableName);
    HTable hTable = null;
    // If the table descriptor is in the cache, we assume the table exists.
    if (!tableDescriptors.containsKey(tableId)) {
        synchronized (this) {
            if (!tableDescriptors.containsKey(tableId)) {
                try (HBaseDDLExecutor ddlExecutor = ddlExecutorFactory.get()) {
                    ColumnFamilyDescriptorBuilder cfdBuilder = HBaseTableUtil.getColumnFamilyDescriptorBuilder(Bytes.toString(COLUMN_FAMILY), hConf);
                    TableDescriptorBuilder tdBuilder = HBaseTableUtil.getTableDescriptorBuilder(tableId, cConf).addColumnFamily(cfdBuilder.build());
                    ddlExecutor.createTableIfNotExists(tdBuilder.build(), null);
                    hTable = tableUtil.createHTable(hConf, tableId);
                    tableDescriptors.put(tableId, hTable.getTableDescriptor());
                }
            }
        }
    }
    if (hTable == null) {
        hTable = tableUtil.createHTable(hConf, tableId);
    }
    return new HBaseMetadataTable(tableUtil, hTable, COLUMN_FAMILY, cConf.getInt(Constants.MessagingSystem.HBASE_SCAN_CACHE_ROWS), createExceptionHandler(tableId));
}
Also used : TableId(co.cask.cdap.data2.util.TableId) HBaseDDLExecutor(co.cask.cdap.spi.hbase.HBaseDDLExecutor) ColumnFamilyDescriptorBuilder(co.cask.cdap.data2.util.hbase.ColumnFamilyDescriptorBuilder) HTableDescriptorBuilder(co.cask.cdap.data2.util.hbase.HTableDescriptorBuilder) TableDescriptorBuilder(co.cask.cdap.data2.util.hbase.TableDescriptorBuilder) HTable(org.apache.hadoop.hbase.client.HTable)

Example 5 with TableDescriptorBuilder

use of co.cask.cdap.data2.util.hbase.TableDescriptorBuilder in project cdap by caskdata.

the class HBaseTableFactory method createTable.

/**
 * Creates a new instance of {@link HTable} for the given {@link TableId}. If the hbase table doesn't
 * exist, a new one will be created with the given number of splits.
 */
private HTableWithRowKeyDistributor createTable(TableId tableId, int splits, Class<? extends Coprocessor> coprocessor) throws IOException {
    // Lookup the table descriptor from the cache first. If it is there, we assume the HBase table exists
    // Otherwise, attempt to create it.
    HTable hTable = null;
    HTableDescriptor htd = tableDescriptors.get(tableId);
    if (htd == null) {
        synchronized (this) {
            htd = tableDescriptors.get(tableId);
            if (htd == null) {
                boolean tableExists;
                try (HBaseAdmin admin = new HBaseAdmin(hConf)) {
                    tableExists = tableUtil.tableExists(admin, tableId);
                }
                // Create the table if the table doesn't exist
                try (HBaseDDLExecutor ddlExecutor = ddlExecutorFactory.get()) {
                    // If table exists, then skip creating coprocessor etc
                    if (!tableExists) {
                        TableId metadataTableId = tableUtil.createHTableId(NamespaceId.SYSTEM, cConf.get(Constants.MessagingSystem.METADATA_TABLE_NAME));
                        ColumnFamilyDescriptorBuilder cfdBuilder = HBaseTableUtil.getColumnFamilyDescriptorBuilder(Bytes.toString(COLUMN_FAMILY), hConf);
                        TableDescriptorBuilder tdBuilder = HBaseTableUtil.getTableDescriptorBuilder(tableId, cConf).addColumnFamily(cfdBuilder.build()).addProperty(Constants.MessagingSystem.HBASE_MESSAGING_TABLE_PREFIX_NUM_BYTES, Integer.toString(1)).addProperty(Constants.MessagingSystem.KEY_DISTRIBUTOR_BUCKETS_ATTR, Integer.toString(splits)).addProperty(Constants.MessagingSystem.HBASE_METADATA_TABLE_NAMESPACE, metadataTableId.getNamespace()).addProperty(HTableDescriptor.SPLIT_POLICY, cConf.get(Constants.MessagingSystem.TABLE_HBASE_SPLIT_POLICY)).addCoprocessor(coprocessorManager.getCoprocessorDescriptor(coprocessor, Coprocessor.PRIORITY_USER));
                        // Set the key distributor size the same as the initial number of splits,
                        // essentially one bucket per split.
                        byte[][] splitKeys = HBaseTableUtil.getSplitKeys(splits, splits, new RowKeyDistributorByHashPrefix(new OneByteSimpleHash(splits)));
                        ddlExecutor.createTableIfNotExists(tdBuilder.build(), splitKeys);
                        hTable = tableUtil.createHTable(hConf, tableId);
                        htd = hTable.getTableDescriptor();
                        tableDescriptors.put(tableId, htd);
                    } else {
                        hTable = tableUtil.createHTable(hConf, tableId);
                        htd = hTable.getTableDescriptor();
                        tableDescriptors.put(tableId, htd);
                    }
                }
            }
        }
    }
    if (hTable == null) {
        hTable = tableUtil.createHTable(hConf, tableId);
    }
    hTable.setAutoFlushTo(false);
    return new HTableWithRowKeyDistributor(hTable, new RowKeyDistributorByHashPrefix(new OneByteSimpleHash(getKeyDistributorBuckets(tableId, htd))));
}
Also used : HBaseDDLExecutor(co.cask.cdap.spi.hbase.HBaseDDLExecutor) TableId(co.cask.cdap.data2.util.TableId) HBaseAdmin(org.apache.hadoop.hbase.client.HBaseAdmin) RowKeyDistributorByHashPrefix(co.cask.cdap.hbase.wd.RowKeyDistributorByHashPrefix) ColumnFamilyDescriptorBuilder(co.cask.cdap.data2.util.hbase.ColumnFamilyDescriptorBuilder) OneByteSimpleHash(co.cask.cdap.hbase.wd.RowKeyDistributorByHashPrefix.OneByteSimpleHash) HTableDescriptorBuilder(co.cask.cdap.data2.util.hbase.HTableDescriptorBuilder) TableDescriptorBuilder(co.cask.cdap.data2.util.hbase.TableDescriptorBuilder) HTable(org.apache.hadoop.hbase.client.HTable) HTableDescriptor(org.apache.hadoop.hbase.HTableDescriptor)

Aggregations

TableId (co.cask.cdap.data2.util.TableId)7 HBaseDDLExecutor (co.cask.cdap.spi.hbase.HBaseDDLExecutor)7 ColumnFamilyDescriptorBuilder (co.cask.cdap.data2.util.hbase.ColumnFamilyDescriptorBuilder)5 TableDescriptorBuilder (co.cask.cdap.data2.util.hbase.TableDescriptorBuilder)5 HTable (org.apache.hadoop.hbase.client.HTable)5 HTableDescriptorBuilder (co.cask.cdap.data2.util.hbase.HTableDescriptorBuilder)3 NamespaceId (co.cask.cdap.proto.id.NamespaceId)3 RowKeyDistributorByHashPrefix (co.cask.cdap.hbase.wd.RowKeyDistributorByHashPrefix)2 IOException (java.io.IOException)2 HBaseAdmin (org.apache.hadoop.hbase.client.HBaseAdmin)2 AbstractRowKeyDistributor (co.cask.cdap.hbase.wd.AbstractRowKeyDistributor)1 OneByteSimpleHash (co.cask.cdap.hbase.wd.RowKeyDistributorByHashPrefix.OneByteSimpleHash)1 VisibleForTesting (com.google.common.annotations.VisibleForTesting)1 Map (java.util.Map)1 HTableDescriptor (org.apache.hadoop.hbase.HTableDescriptor)1 Delete (org.apache.hadoop.hbase.client.Delete)1 Put (org.apache.hadoop.hbase.client.Put)1