Field Detail

• LOG

  private static final org.apache.commons.logging.Log LOG

• CLIENT_MAX_THREADS

  public static final String CLIENT_MAX_THREADS

  See Also:

  Constant Field Values

• USE_NATIVE_TRANSPORT

  public static final String USE_NATIVE_TRANSPORT

  See Also:

  Constant Field Values

• USE_GLOBAL_EVENT_LOOP_GROUP

  public static final String USE_GLOBAL_EVENT_LOOP_GROUP

  See Also:

  Constant Field Values

• WHEEL_TIMER

  private static final io.netty.util.HashedWheelTimer WHEEL_TIMER

• DEFAULT_CHANNEL_INITIALIZER

  private static final io.netty.channel.ChannelInitializer<io.netty.channel.socket.SocketChannel> DEFAULT_CHANNEL_INITIALIZER

• callIdCnt

  protected final AtomicInteger callIdCnt

• connections

  private final PoolMap<Integer,AsyncRpcChannel> connections

• failedServers

  final FailedServers failedServers

• bootstrap

  final io.netty.bootstrap.Bootstrap bootstrap

• useGlobalEventLoopGroup

  private final boolean useGlobalEventLoopGroup

• GLOBAL_EVENT_LOOP_GROUP

  static Pair<io.netty.channel.EventLoopGroup,Class<? extends io.netty.channel.Channel>> GLOBAL_EVENT_LOOP_GROUP

• closed

  private boolean closed

Constructor Detail

• AsyncRpcClient

  protected AsyncRpcClient(org.apache.hadoop.conf.Configuration configuration,
                String clusterId,
                SocketAddress localAddress,
                MetricsConnection metrics,
                io.netty.channel.ChannelInitializer<io.netty.channel.socket.SocketChannel> channelInitializer)

  Constructor for tests

  Parameters:

  configuration - to HBase

  clusterId - for the cluster

  localAddress - local address to connect to

  metrics - the connection metrics

  channelInitializer - for custom channel handlers

• AsyncRpcClient

  AsyncRpcClient(org.apache.hadoop.conf.Configuration configuration)

  Used in test only.

• AsyncRpcClient

  AsyncRpcClient(org.apache.hadoop.conf.Configuration configuration,
                io.netty.channel.ChannelInitializer<io.netty.channel.socket.SocketChannel> channelInitializer)

  Used in test only.

• AsyncRpcClient

  public AsyncRpcClient(org.apache.hadoop.conf.Configuration configuration,
                String clusterId,
                SocketAddress localAddress,
                MetricsConnection metrics)

  Constructor

  Parameters:

  configuration - to HBase

  clusterId - for the cluster

  localAddress - local address to connect to

  metrics - the connection metrics

Method Detail

• getGlobalEventLoopGroup

  private static Pair<io.netty.channel.EventLoopGroup,Class<? extends io.netty.channel.Channel>> getGlobalEventLoopGroup(org.apache.hadoop.conf.Configuration conf)

• createEventLoopGroup

  private static Pair<io.netty.channel.EventLoopGroup,Class<? extends io.netty.channel.Channel>> createEventLoopGroup(org.apache.hadoop.conf.Configuration conf)

• call

  protected Pair<com.google.protobuf.Message,CellScanner> call(PayloadCarryingRpcController pcrc,
                                                   com.google.protobuf.Descriptors.MethodDescriptor md,
                                                   com.google.protobuf.Message param,
                                                   com.google.protobuf.Message returnType,
                                                   User ticket,
                                                   InetSocketAddress addr,
                                                   MetricsConnection.CallStats callStats)
                                                        throws IOException,
                                                               InterruptedException

  Make a call, passing param, to the IPC server running at address which is servicing the protocol protocol, with the ticket credentials, returning the value. Throws exceptions if there are network problems or if the remote code threw an exception.

  Specified by:

  call in class AbstractRpcClient

  Parameters:

  ticket - Be careful which ticket you pass. A new user will mean a new Connection. UserProvider.getCurrent() makes a new instance of User each time so will be a new Connection each time.

  Returns:

  A pair with the Message response and the Cell data (if any).

  Throws:

  InterruptedException - if call is interrupted

  IOException - if a connection failure is encountered

• callMethod

  private void callMethod(com.google.protobuf.Descriptors.MethodDescriptor md,
                PayloadCarryingRpcController pcrc,
                com.google.protobuf.Message param,
                com.google.protobuf.Message returnType,
                User ticket,
                InetSocketAddress addr,
                com.google.protobuf.RpcCallback<com.google.protobuf.Message> done)

  Call method async

• close

  public void close()

  Close netty

• createCellScanner

  public CellScanner createCellScanner(byte[] cellBlock)
                                throws IOException

  Create a cell scanner

  Parameters:

  cellBlock - to create scanner for

  Returns:

  CellScanner

  Throws:

  IOException - on error on creation cell scanner

• buildCellBlock

  public ByteBuffer buildCellBlock(CellScanner cells)
                            throws IOException

  Build cell block

  Parameters:

  cells - to create block with

  Returns:

  ByteBuffer with cells

  Throws:

  IOException - if block creation fails

• createRpcChannel

  private AsyncRpcChannel createRpcChannel(String serviceName,
                                 InetSocketAddress location,
                                 User ticket)
                                    throws StoppedRpcClientException,
                                           FailedServerException

  Creates an RPC client

  Parameters:

  serviceName - name of servicce

  location - to connect to

  ticket - for current user

  Returns:

  new RpcChannel

  Throws:

  StoppedRpcClientException - when Rpc client is stopped

  FailedServerException - if server failed

• cancelConnections

  public void cancelConnections(ServerName sn)

  Interrupt the connections to the given ip:port server. This should be called if the server is known as actually dead. This will not prevent current operation to be retried, and, depending on their own behavior, they may retry on the same server. This can be a feature, for example at startup. In any case, they're likely to get connection refused (if the process died) or no route to host: i.e. there next retries should be faster and with a safe exception.

  Parameters:

  sn - server to cancel connections for

• removeConnection

  public void removeConnection(AsyncRpcChannel connection)

  Remove connection from pool

• createRpcChannel

  public com.google.protobuf.RpcChannel createRpcChannel(ServerName sn,
                                                User user,
                                                int rpcTimeout)

  Creates a "channel" that can be used by a protobuf service. Useful setting up protobuf stubs.

  Parameters:

  sn - server name describing location of server

  user - which is to use the connection

  rpcTimeout - default rpc operation timeout

  Returns:

  A rpc channel that goes via this rpc client instance.

• newTimeout

  io.netty.util.Timeout newTimeout(io.netty.util.TimerTask task,
                                 long delay,
                                 TimeUnit unit)