Using TCP

Connecting

public class Client extends UntypedActor {
  
  final InetSocketAddress remote;
  final ActorRef listener;
  
  public static Props props(InetSocketAddress remote, ActorRef listener) {
      return Props.create(Client.class, remote, listener);
  }

  public Client(InetSocketAddress remote, ActorRef listener) {
    this.remote = remote;
    this.listener = listener;
    
    final ActorRef tcp = Tcp.get(getContext().system()).manager();
    tcp.tell(TcpMessage.connect(remote), getSelf());
  }

  @Override
  public void onReceive(Object msg) throws Exception {
    if (msg instanceof CommandFailed) {
      listener.tell("failed", getSelf());
      getContext().stop(getSelf());
      
    } else if (msg instanceof Connected) {
      listener.tell(msg, getSelf());
      getSender().tell(TcpMessage.register(getSelf()), getSelf());
      getContext().become(connected(getSender()));
    }
  }

  private Procedure<Object> connected(final ActorRef connection) {
    return new Procedure<Object>() {
      @Override
      public void apply(Object msg) throws Exception {
        
        if (msg instanceof ByteString) {
          connection.tell(TcpMessage.write((ByteString) msg), getSelf());
      
        } else if (msg instanceof CommandFailed) {
          // OS kernel socket buffer was full
        
        } else if (msg instanceof Received) {
          listener.tell(((Received) msg).data(), getSelf());
        
        } else if (msg.equals("close")) {
          connection.tell(TcpMessage.close(), getSelf());
        
        } else if (msg instanceof ConnectionClosed) {
          getContext().stop(getSelf());
        }
      }
    };
  }
  
}

The first step of connecting to a remote address is sending a Connect message to the TCP manager; in addition to the simplest form shown above there is also the possibility to specify a local InetSocketAddress to bind to and a list of socket options to apply.

The TCP manager will then reply either with a CommandFailed or it will spawn an internal actor representing the new connection. This new actor will then send a Connected message to the original sender of the Connect message.

In order to activate the new connection a Register message must be sent to the connection actor, informing that one about who shall receive data from the socket. Before this step is done the connection cannot be used, and there is an internal timeout after which the connection actor will shut itself down if no Register message is received.

The connection actor watches the registered handler and closes the connection when that one terminates, thereby cleaning up all internal resources associated with that connection.

The actor in the example above uses become to switch from unconnected to connected operation, demonstrating the commands and events which are observed in that state. For a discussion on CommandFailed see Throttling Reads and Writes below. ConnectionClosed is a trait, which marks the different connection close events. The last line handles all connection close events in the same way. It is possible to listen for more fine-grained connection close events, see Closing Connections below.

Accepting connections

public class Server extends UntypedActor {
  
  final ActorRef manager;
  
  public Server(ActorRef manager) {
    this.manager = manager;
  }
  
  public static Props props(ActorRef manager) {
    return Props.create(Server.class, manager);
  }

  @Override
  public void preStart() throws Exception {
    final ActorRef tcp = Tcp.get(getContext().system()).manager();
    tcp.tell(TcpMessage.bind(getSelf(),
        new InetSocketAddress("localhost", 0), 100), getSelf());
  }

  @Override
  public void onReceive(Object msg) throws Exception {
    if (msg instanceof Bound) {
      manager.tell(msg, getSelf());

    } else if (msg instanceof CommandFailed) {
      getContext().stop(getSelf());
    
    } else if (msg instanceof Connected) {
      final Connected conn = (Connected) msg;
      manager.tell(conn, getSelf());
      final ActorRef handler = getContext().actorOf(
          Props.create(SimplisticHandler.class));
      getSender().tell(TcpMessage.register(handler), getSelf());
    }
  }
  
}

To create a TCP server and listen for inbound connections, a Bind command has to be sent to the TCP manager. This will instruct the TCP manager to listen for TCP connections on a particular InetSocketAddress; the port may be specified as 0 in order to bind to a random port.

The actor sending the Bind message will receive a Bound message signaling that the server is ready to accept incoming connections; this message also contains the InetSocketAddress to which the socket was actually bound (i.e. resolved IP address and correct port number).

From this point forward the process of handling connections is the same as for outgoing connections. The example demonstrates that handling the reads from a certain connection can be delegated to another actor by naming it as the handler when sending the Register message. Writes can be sent from any actor in the system to the connection actor (i.e. the actor which sent the Connected message). The simplistic handler is defined as:

public class SimplisticHandler extends UntypedActor {
  @Override
  public void onReceive(Object msg) throws Exception {
    if (msg instanceof Received) {
      final ByteString data = ((Received) msg).data();
      System.out.println(data);
      getSender().tell(TcpMessage.write(data), getSelf());
    } else if (msg instanceof ConnectionClosed) {
      getContext().stop(getSelf());
    }
  }
}

For a more complete sample which also takes into account the possibility of failures when sending please see Throttling Reads and Writes below.

The only difference to outgoing connections is that the internal actor managing the listen port—the sender of the Bound message—watches the actor which was named as the recipient for Connected messages in the Bind message. When that actor terminates the listen port will be closed and all resources associated with it will be released; existing connections will not be terminated at this point.

Closing connections

A connection can be closed by sending one of the commands Close, ConfirmedClose or Abort to the connection actor.

Close will close the connection by sending a FIN message, but without waiting for confirmation from the remote endpoint. Pending writes will be flushed. If the close is successful, the listener will be notified with Closed.

ConfirmedClose will close the sending direction of the connection by sending a FIN message, but data will continue to be received until the remote endpoint closes the connection, too. Pending writes will be flushed. If the close is successful, the listener will be notified with ConfirmedClosed.

Abort will immediately terminate the connection by sending a RST message to the remote endpoint. Pending writes will be not flushed. If the close is successful, the listener will be notified with Aborted.

PeerClosed will be sent to the listener if the connection has been closed by the remote endpoint. Per default, the connection will then automatically be closed from this endpoint as well. To support half-closed connections set the keepOpenOnPeerClosed member of the Register message to true in which case the connection stays open until it receives one of the above close commands.

ErrorClosed will be sent to the listener whenever an error happened that forced the connection to be closed.

All close notifications are sub-types of ConnectionClosed so listeners who do not need fine-grained close events may handle all close events in the same way.

Writing to a connection

Once a connection has been established data can be sent to it from any actor in the form of a Tcp.WriteCommand. Tcp.WriteCommand is an abstract class with three concrete implementations:

Tcp.Write

The simplest WriteCommand implementation which wraps a ByteString instance and an "ack" event. A ByteString (as explained in this section) models one or more chunks of immutable in-memory data with a maximum (total) size of 2 GB (2^31 bytes).

Tcp.WriteFile

If you want to send "raw" data from a file you can do so efficiently with the Tcp.WriteFile command. This allows you do designate a (contiguous) chunk of on-disk bytes for sending across the connection without the need to first load them into the JVM memory. As such Tcp.WriteFile can "hold" more than 2GB of data and an "ack" event if required.

Tcp.CompoundWrite

Sometimes you might want to group (or interleave) several Tcp.Write and/or Tcp.WriteFile commands into one atomic write command which gets written to the connection in one go. The Tcp.CompoundWrite allows you to do just that and offers three benefits:

1. As explained in the following section the TCP connection actor can only handle one single write command at a time. By combining several writes into one CompoundWrite you can have them be sent across the connection with minimum overhead and without the need to spoon feed them to the connection actor via an ACK-based message protocol.
2. Because a WriteCommand is atomic you can be sure that no other actor can "inject" other writes into your series of writes if you combine them into one single CompoundWrite. In scenarios where several actors write to the same connection this can be an important feature which can be somewhat hard to achieve otherwise.
3. The "sub writes" of a CompoundWrite are regular Write or WriteFile commands that themselves can request "ack" events. These ACKs are sent out as soon as the respective "sub write" has been completed. This allows you to attach more than one ACK to a Write or WriteFile (by combining it with an empty write that itself requests an ACK) or to have the connection actor acknowledge the progress of transmitting the CompoundWrite by sending out intermediate ACKs at arbitrary points.

Throttling Reads and Writes

The basic model of the TCP connection actor is that it has no internal buffering (i.e. it can only process one write at a time, meaning it can buffer one write until it has been passed on to the O/S kernel in full). Congestion needs to be handled at the user level, for both writes and reads.

For back-pressuring writes there are three modes of operation

• ACK-based: every Write command carries an arbitrary object, and if this object is not Tcp.NoAck then it will be returned to the sender of the Write upon successfully writing all contained data to the socket. If no other write is initiated before having received this acknowledgement then no failures can happen due to buffer overrun.
• NACK-based: every write which arrives while a previous write is not yet completed will be replied to with a CommandFailed message containing the failed write. Just relying on this mechanism requires the implemented protocol to tolerate skipping writes (e.g. if each write is a valid message on its own and it is not required that all are delivered). This mode is enabled by setting the useResumeWriting flag to false within the Register message during connection activation.
• NACK-based with write suspending: this mode is very similar to the NACK-based one, but once a single write has failed no further writes will succeed until a ResumeWriting message is received. This message will be answered with a WritingResumed message once the last accepted write has completed. If the actor driving the connection implements buffering and resends the NACK’ed messages after having awaited the WritingResumed signal then every message is delivered exactly once to the network socket.

These write models (with the exception of the second which is rather specialised) are demonstrated in complete examples below. The full and contiguous source is available on GitHub.

For back-pressuring reads there are two modes of operation

• Push-reading: in this mode the connection actor sends the registered reader actor incoming data as soon as available as Received events. Whenever the reader actor wants to signal back-pressure to the remote TCP endpoint it can send a SuspendReading message to the connection actor to indicate that it wants to suspend the reception of new data. No Received events will arrive until a corresponding ResumeReading is sent indicating that the receiver actor is ready again.
• Pull-reading: after sending a Received event the connection actor automatically suspends accepting data from the socket until the reader actor signals with a ResumeReading message that it is ready to process more input data. Hence new data is "pulled" from the connection by sending ResumeReading messages.

ACK-Based Write Back-Pressure

For proper function of the following example it is important to configure the connection to remain half-open when the remote side closed its writing end: this allows the example EchoHandler to write all outstanding data back to the client before fully closing the connection. This is enabled using a flag upon connection activation (observe the Register message):

connection.tell(TcpMessage.register(handler,
    true, // <-- keepOpenOnPeerClosed flag
    true), getSelf());

With this preparation let us dive into the handler itself:

public class SimpleEchoHandler extends UntypedActor {
  
  final LoggingAdapter log = Logging
      .getLogger(getContext().system(), getSelf());

  final ActorRef connection;
  final InetSocketAddress remote;

  public static final long maxStored = 100000000;
  public static final long highWatermark = maxStored * 5 / 10;
  public static final long lowWatermark = maxStored * 2 / 10;

  public SimpleEchoHandler(ActorRef connection, InetSocketAddress remote) {
    this.connection = connection;
    this.remote = remote;

    // sign death pact: this actor stops when the connection is closed
    getContext().watch(connection);
  }

  @Override
  public void onReceive(Object msg) throws Exception {
    if (msg instanceof Received) {
      final ByteString data = ((Received) msg).data();
      buffer(data);
      connection.tell(TcpMessage.write(data, ACK), getSelf());
      // now switch behavior to “waiting for acknowledgement”
      getContext().become(buffering, false);

    } else if (msg instanceof ConnectionClosed) {
      getContext().stop(getSelf());
    }
  }

  private final Procedure<Object> buffering = new Procedure<Object>() {
    @Override
    public void apply(Object msg) throws Exception {
      if (msg instanceof Received) {
        buffer(((Received) msg).data());
        
      } else if (msg == ACK) {
        acknowledge();

      } else if (msg instanceof ConnectionClosed) {
        if (((ConnectionClosed) msg).isPeerClosed()) {
          closing = true;
        } else {
          // could also be ErrorClosed, in which case we just give up
          getContext().stop(getSelf());
        }
      }
    }
  };

  // storage omitted ...
}

The principle is simple: when having written a chunk always wait for the Ack to come back before sending the next chunk. While waiting we switch behavior such that new incoming data are buffered. The helper functions used are a bit lengthy but not complicated:

protected void buffer(ByteString data) {
  storage.add(data);
  stored += data.size();

  if (stored > maxStored) {
    log.warning("drop connection to [{}] (buffer overrun)", remote);
    getContext().stop(getSelf());

  } else if (stored > highWatermark) {
    log.debug("suspending reading");
    connection.tell(TcpMessage.suspendReading(), getSelf());
    suspended = true;
  }
}

protected void acknowledge() {
  final ByteString acked = storage.remove();
  stored -= acked.size();
  transferred += acked.size();

  if (suspended && stored < lowWatermark) {
    log.debug("resuming reading");
    connection.tell(TcpMessage.resumeReading(), getSelf());
    suspended = false;
  }
  
  if (storage.isEmpty()) {
    if (closing) {
      getContext().stop(getSelf());
    } else {
      getContext().unbecome();
    }
  } else {
    connection.tell(TcpMessage.write(storage.peek(), ACK), getSelf());
  }
}

The most interesting part is probably the last: an Ack removes the oldest data chunk from the buffer, and if that was the last chunk then we either close the connection (if the peer closed its half already) or return to the idle behavior; otherwise we just send the next buffered chunk and stay waiting for the next Ack.

Back-pressure can be propagated also across the reading side back to the writer on the other end of the connection by sending the SuspendReading command to the connection actor. This will lead to no data being read from the socket anymore (although this does happen after a delay because it takes some time until the connection actor processes this command, hence appropriate head-room in the buffer should be present), which in turn will lead to the O/S kernel buffer filling up on our end, then the TCP window mechanism will stop the remote side from writing, filling up its write buffer, until finally the writer on the other side cannot push any data into the socket anymore. This is how end-to-end back-pressure is realized across a TCP connection.

NACK-Based Write Back-Pressure with Suspending

public class EchoHandler extends UntypedActor {

  final LoggingAdapter log = Logging
      .getLogger(getContext().system(), getSelf());

  final ActorRef connection;
  final InetSocketAddress remote;

  public static final long MAX_STORED = 100000000;
  public static final long HIGH_WATERMARK = MAX_STORED * 5 / 10;
  public static final long LOW_WATERMARK = MAX_STORED * 2 / 10;
  
  private static class Ack implements Event {
    public final int ack;
    public Ack(int ack) {
      this.ack = ack;
    }
  }

  public EchoHandler(ActorRef connection, InetSocketAddress remote) {
    this.connection = connection;
    this.remote = remote;

    // sign death pact: this actor stops when the connection is closed
    getContext().watch(connection);

    // start out in optimistic write-through mode
    getContext().become(writing);
  }

  private final Procedure<Object> writing = new Procedure<Object>() {
    @Override
    public void apply(Object msg) throws Exception {
      if (msg instanceof Received) {
        final ByteString data = ((Received) msg).data();
        connection.tell(TcpMessage.write(data, new Ack(currentOffset())), getSelf());
        buffer(data);

      } else if (msg instanceof Integer) {
        acknowledge((Integer) msg);

      } else if (msg instanceof CommandFailed) {
        final Write w = (Write) ((CommandFailed) msg).cmd();
        connection.tell(TcpMessage.resumeWriting(), getSelf());
        getContext().become(buffering((Ack) w.ack()));

      } else if (msg instanceof ConnectionClosed) {
        final ConnectionClosed cl = (ConnectionClosed) msg;
        if (cl.isPeerClosed()) {
          if (storage.isEmpty()) {
            getContext().stop(getSelf());
          } else {
            getContext().become(closing);
          }
        }
      }
    }
  };

  // buffering ...

  // closing ...

  // storage omitted ...
}

The principle here is to keep writing until a CommandFailed is received, using acknowledgements only to prune the resend buffer. When a such a failure was received, transition into a different state for handling and handle resending of all queued data:

protected Procedure<Object> buffering(final Ack nack) {
  return new Procedure<Object>() {

    private int toAck = 10;
    private boolean peerClosed = false;

    @Override
    public void apply(Object msg) throws Exception {
      if (msg instanceof Received) {
        buffer(((Received) msg).data());

      } else if (msg instanceof WritingResumed) {
        writeFirst();

      } else if (msg instanceof ConnectionClosed) {
        if (((ConnectionClosed) msg).isPeerClosed())
          peerClosed = true;
        else
          getContext().stop(getSelf());

      } else if (msg instanceof Integer) {
        final int ack = (Integer) msg;
        acknowledge(ack);

        if (ack >= nack.ack) {
          // otherwise it was the ack of the last successful write

          if (storage.isEmpty()) {
            if (peerClosed)
              getContext().stop(getSelf());
            else
              getContext().become(writing);

          } else {
            if (toAck > 0) {
              // stay in ACK-based mode for a short while
              writeFirst();
              --toAck;
            } else {
              // then return to NACK-based again
              writeAll();
              if (peerClosed)
                getContext().become(closing);
              else
                getContext().become(writing);
            }
          }
        }
      }
    }
  };
}

It should be noted that all writes which are currently buffered have also been sent to the connection actor upon entering this state, which means that the ResumeWriting message is enqueued after those writes, leading to the reception of all outstanding CommandFailed messages (which are ignored in this state) before receiving the WritingResumed signal. That latter message is sent by the connection actor only once the internally queued write has been fully completed, meaning that a subsequent write will not fail. This is exploited by the EchoHandler to switch to an ACK-based approach for the first ten writes after a failure before resuming the optimistic write-through behavior.

protected Procedure<Object> closing = new Procedure<Object>() {
  @Override
  public void apply(Object msg) throws Exception {
    if (msg instanceof CommandFailed) {
      // the command can only have been a Write
      connection.tell(TcpMessage.resumeWriting(), getSelf());
      getContext().become(closeResend, false);
    } else if (msg instanceof Integer) {
      acknowledge((Integer) msg);
      if (storage.isEmpty())
        getContext().stop(getSelf());
    }
  }
};

protected Procedure<Object> closeResend = new Procedure<Object>() {
  @Override
  public void apply(Object msg) throws Exception {
    if (msg instanceof WritingResumed) {
      writeAll();
      getContext().unbecome();
    } else if (msg instanceof Integer) {
      acknowledge((Integer) msg);
    }
  }
};

Closing the connection while still sending all data is a bit more involved than in the ACK-based approach: the idea is to always send all outstanding messages and acknowledge all successful writes, and if a failure happens then switch behavior to await the WritingResumed event and start over.

The helper functions are very similar to the ACK-based case:

protected void buffer(ByteString data) {
  storage.add(data);
  stored += data.size();

  if (stored > MAX_STORED) {
    log.warning("drop connection to [{}] (buffer overrun)", remote);
    getContext().stop(getSelf());

  } else if (stored > HIGH_WATERMARK) {
    log.debug("suspending reading at {}", currentOffset());
    connection.tell(TcpMessage.suspendReading(), getSelf());
    suspended = true;
  }
}

protected void acknowledge(int ack) {
  assert ack == storageOffset;
  assert !storage.isEmpty();

  final ByteString acked = storage.remove();
  stored -= acked.size();
  transferred += acked.size();
  storageOffset += 1;

  if (suspended && stored < LOW_WATERMARK) {
    log.debug("resuming reading");
    connection.tell(TcpMessage.resumeReading(), getSelf());
    suspended = false;
  }
}

Read Back-Pressure with Pull Mode

When using push based reading, data coming from the socket is sent to the actor as soon as it is available. In the case of the previous Echo server example this meant that we needed to maintain a buffer of incoming data to keep it around since the rate of writing might be slower than the rate of the arrival of new data.

With the Pull mode this buffer can be completely eliminated as the following snippet demonstrates:

@Override
public void preStart() throws Exception {
    connection.tell(TcpMessage.resumeReading(), getSelf());
}

@Override
public void onReceive(Object message) throws Exception {
    if (message instanceof Tcp.Received) {
        ByteString data = ((Tcp.Received) message).data();
        connection.tell(TcpMessage.write(data, new Ack()), getSelf());
    } else if (message instanceof Ack) {
        connection.tell(TcpMessage.resumeReading(), getSelf());
    }
}

The idea here is that reading is not resumed until the previous write has been completely acknowledged by the connection actor. Every pull mode connection actor starts from suspended state. To start the flow of data we send a ResumeReading in the preStart method to tell the connection actor that we are ready to receive the first chunk of data. Since we only resume reading when the previous data chunk has been completely written there is no need for maintaining a buffer.

To enable pull reading on an outbound connection the pullMode parameter of the Connect should be set to true:

final List<Inet.SocketOption> options = new ArrayList<Inet.SocketOption>();
tcp.tell(
   TcpMessage.connect(new InetSocketAddress("localhost", 3000), null, options, null, true),
   getSelf()
);

tcp = Tcp.get(getContext().system()).manager();
final List<Inet.SocketOption> options = new ArrayList<Inet.SocketOption>();
tcp.tell(
   TcpMessage.bind(getSelf(), new InetSocketAddress("localhost", 0), 100, options, true),
   getSelf()
);

public void onReceive(Object message) throws Exception {
    if (message instanceof Tcp.Bound) {
        listener = getSender();
        // Accept connections one by one
        listener.tell(TcpMessage.resumeAccepting(1), getSelf());
    } else if (message instanceof Tcp.Connected) {
        ActorRef handler = getContext().actorOf(Props.create(PullEcho.class, getSender()));
        getSender().tell(TcpMessage.register(handler), getSelf());
        // Resume accepting connections
        listener.tell(TcpMessage.resumeAccepting(1), getSelf());
    }
}