Hadoop中客户端和服务器端的方法调用过程

1、Java动态代理实例

Java 动态代理一个简单的demo：（用以对比Hadoop中的动态代理）

Hello接口：

public interface Hello {
      void sayHello(String to);
      void print(String p);
}

Hello接口的实现类：

public class HelloImpl implements Hello { 
     
   public void sayHello(String to) { 
        System.out.println("Say hello to " + to); 
    } 
     
   public void print(String s) { 
        System.out.println("print : " + s); 
    } 
     
}

与代理类（HelloImpl类）相关联的InvocationHandler对象

public class LogHandler implements InvocationHandler { 
     
   private Object dele; 
     
   public LogHandler(Object obj) { 
       this.dele = obj; 
    } 
     
   public Object invoke(Object proxy, Method method, Object[] args) throws Throwable { 
        doBefore(); 
       //在这里完全可以把下面这句注释掉，而做一些其它的事情 

        Object result = method.invoke(dele, args); 
        after(); 
       return result; 
    } 
     
   private void doBefore() { 
        System.out.println("before...."); 
    } 
     
   private void after() { 
        System.out.println("after...."); 
    } 
}

最后测试代码如下：

public class ProxyTest { 
 
   public static void main(String[] args) { 
        HelloImpl impl = new HelloImpl(); 
        LogHandler handler= new LogHandler(impl); 
       //这里把handler与impl新生成的代理类相关联 

        Hello hello = (Hello) Proxy.newProxyInstance(impl.getClass().getClassLoader(), impl.getClass().getInterfaces(), handler); 
         
       //这里无论访问哪个方法，都是会把请求转发到handler.invoke

        hello.print("All the test"); 
        hello.sayHello("Denny"); 
    } 
 
}

 

2、Hadoop中的动态代理

2.1、客户端方法调用过程

IPC客户端的处理比动态代理实例稍微复杂：代理对象上的调用被InvocationHandler捕获后，请求被打包并通过IPC连接发送到服务器上，客户端等待并在服务器的处理应答到达后，生成并返回调用结果。IPC上的调用是个同步操作，即，线程会一直等待调用结束，才会开始后续处理；而网络的处理时异步的，请求发送后，不需要等待应答。客户端通过java的wait()/notify()机制简单地解决了异步网络处理和同步IPC调用的差异。

 

Hadoop对外提供查询文件状态的接口，如下：

public interface IPCQueryStatus extends VersionedProtocol {
    IPCFileStatus getFileStatus(String filename);
}

客户端通过如下代码调用：

IPCQueryStatus query = (IPCQueryStatus) RPC.getProxy(IPCQueryStatus.class, IPCQueryServer.IPC_VER, addr, new Configuration());
IPCFileStatus status = query.getFileStatus("\tmp\testIPC");

2.1.1、Client端动态代理实现

在RPC的getProxy代码如下：

public static VersionedProtocol getProxy(
      Class<? extends VersionedProtocol> protocol,
      long clientVersion, InetSocketAddress addr, UserGroupInformation ticket,
      Configuration conf, SocketFactory factory, int rpcTimeout) throws IOException {

    ......
    VersionedProtocol proxy =
        (VersionedProtocol) Proxy.newProxyInstance(
            protocol.getClassLoader(), new Class[] { protocol },
            new Invoker(protocol, addr, ticket, conf, factory, rpcTimeout));
    ......
    return proxy;
    ......
  }

需要制定一个InvocationHandler，对于所有的调用请求，这个InvocationHandler都是Invoke，如下：

private static class Invoker implements InvocationHandler {
    private Client.ConnectionId remoteId;// 用来标示一个connection，用以复用
    private Client client;//最重要的成员变量，RPC客户端
    private boolean isClosed = false;

    public Invoker(Class<? extends VersionedProtocol> protocol,
        InetSocketAddress address, UserGroupInformation ticket,
        Configuration conf, SocketFactory factory,
        int rpcTimeout) throws IOException {
      this.remoteId = Client.ConnectionId.getConnectionId(address, protocol,
          ticket, rpcTimeout, conf);
      this.client = CLIENTS.getClient(conf, factory);//★
    }
    ......

    public Object invoke(Object proxy, Method method, Object[] args)
      ......

      ObjectWritable value = (ObjectWritable)
        client.call(new Invocation(method, args), remoteId);
      ......

      return value.get();
    }
}

在上面的代码中，client负责发送IPC请求，并获取结果，类似最上面demo中LogHandler中的dele。

2.1.2、Client通过Connection发送IPC请求并获取结果

如下为client.call方法调用Connection.sendParam发送IPC请求：

public Writable call(Writable param, ConnectionId remoteId)
                       throws InterruptedException, IOException {
    Call call = new Call(param);
    Connection connection = getConnection(remoteId, call);
    connection.sendParam(call);                 // send the parameter
    ...
    synchronized (call) {
      while (!call.done) {
        try {
          call.wait();                           // wait for the result
        } catch (InterruptedException ie) {
          ...
        }
      }

      ...
      if (call.error != null) {
        ...
        throw call.error;
        ...
      } else {
        return call.value;
      }
    }
}

connection.sendParam后，会再调用receiveMessage来获取返回结果。如下：

private class Connection extends Thread {
    ......

    public void run() {
        ......
        while (waitForWork()) {//wait here for work - read or close connection
        receiveResponse();
        }
        ......
    }
    ......
    private void receiveResponse() {
      ......
      touch();

      try {
        int id = in.readInt();                    // try to read an id
        ......
        Call call = calls.get(id);

        int state = in.readInt();     // read call status
        if (state == Status.SUCCESS.state) {
          Writable value = ReflectionUtils.newInstance(valueClass, conf);
          value.readFields(in);                 // read value
          call.setValue(value);
          calls.remove(id);
        } else if (state == Status.ERROR.state) {
          call.setException(new RemoteException(WritableUtils.readString(in),
                                                WritableUtils.readString(in)));
          calls.remove(id);
        } else if (state == Status.FATAL.state) {
          // Close the connection
          markClosed(new RemoteException(WritableUtils.readString(in),
                                         WritableUtils.readString(in)));
        }
      } catch (IOException e) {
        markClosed(e);
      }
    }
}

connection会调用call的setValue或者setException，两个方法都会调用callComplete方法，来调用notify通知进程IPC调用已结束

protected synchronized void callComplete() {
      this.done = true;
      notify();                                 // notify caller
    }

    public synchronized void setException(IOException error) {
      this.error = error;
      callComplete();
    }

    public synchronized void setValue(Writable value) {
      this.value = value;
      callComplete();
    }

 

2.2、服务器端方法调用过程

服务端由Listener接收。

2.2.1、Listener接收IPC请求的工作过程

Listener主要运行NIO选择器循环，并在Listener.doRead()方法中读取数据，Connection.readAndProcess()中恢复数据帧，然后调用processData().

void Listener.doRead(SelectionKey key) throws InterruptedException {
    int count = 0;
    Connection c = (Connection)key.attachment();
    ...
    count = c.readAndProcess();
    ...

}

public int Connection.readAndProcess() throws IOException, InterruptedException {
    ......
    processOneRpc(data.array());
    ......
}

private void Connection.processOneRpc(byte[] buf) throws IOException,
        InterruptedException {
    if (headerRead) {
        processData(buf);
    } else {
        processHeader(buf);
        ......
    }
}

private void Connection.processData(byte[] buf) throws  IOException, InterruptedException {
    DataInputStream dis =
        new DataInputStream(new ByteArrayInputStream(buf));
    int id = dis.readInt();                    // try to read an id

    ......
    Writable param = ReflectionUtils.newInstance(paramClass, conf);//★??paramClass在哪儿设置的★在RPC.Server中，paramClass是Invocation，IPC调用传递的都是Invocation
    param.readFields(dis);        

    Call call = new Call(id, param, this);
    callQueue.put(call);              // queue the call; maybe blocked here
}

ProcessData反序列化调用参数，构造服务器端的Call对象。然后放入callQueue队列中。callQueue阻塞队列定义于Server类中，是Listener和Handler的边界。（生产者Listener消费者Handler）。

2.2.2、Handler处理IPC请求的工作过程

Handler主要工作都在run方法中完成。主循环中，每循环一次处理一个请求（通过调用Server的抽象方法call来完成）。

public void run() {
    ......
    SERVER.set(Server.this);
    ByteArrayOutputStream buf =
    new ByteArrayOutputStream(INITIAL_RESP_BUF_SIZE);
    while (running) {

        final Call call = callQueue.take(); // 获取一个IPC调用
        ......
        String errorClass = null;
        String error = null;
        Writable value = null;

        CurCall.set(call);
        ......
        value = call(call.connection.protocol, call.param,
                   call.timestamp);//实际代码用到jaas，这里简化
        ......

        CurCall.set(null);
        synchronized (call.connection.responseQueue) {
        ......
        setupResponse(buf, call,
                    (error == null) ? Status.SUCCESS : Status.ERROR,
                    value, errorClass, error);
        ...
        responder.doRespond(call);//★?
        }

    }

}

Server.call调用后返回一个writable对象--value，然后通过调用setupResponse将结果序列化到call的Response成员变量中。

private void setupResponse(ByteArrayOutputStream response,
                             Call call, Status status,
                             Writable rv, String errorClass, String error)
  throws IOException {
    response.reset();
    DataOutputStream out = new DataOutputStream(response);
    out.writeInt(call.id);                // write call id
    out.writeInt(status.state);           // write status

    if (status == Status.SUCCESS) {
      rv.write(out);
    } else {
      WritableUtils.writeString(out, errorClass);
      WritableUtils.writeString(out, error);
    }
   ......
    call.setResponse(ByteBuffer.wrap(response.toByteArray()));
  }

Server.call抽象方法的具体实现在RPC.Server中。代码如下：

private Object instance;
......

public Writable call(Class<?> protocol, Writable param, long receivedTime)
    throws IOException {

    Invocation call = (Invocation)param;

    Method method =
      protocol.getMethod(call.getMethodName(),
                               call.getParameterClasses());
    method.setAccessible(true);

    Object value = method.invoke(instance, call.getParameters());

    return new ObjectWritable(method.getReturnType(), value);      

}

Handler所在线程是共享资源，当有一个IPC请求处理完后，即调用Response的doResponse返回结果，而不亲自返回，原因有二：

1. 对共享资源的占用时间越短越好；

2. IPC返回受网络通信时间影响，可能会占用很长时间。

2.2.3、Response的工作过程

doResponse的代码很简单，将Call放入IPC连接的应答队列中，如果应答队列为1，立即调用processResponse发放向客户端发送结果，（队列为1，表明此IPC连接比较空闲，直接发送，避免从Handler线程到Response线程的切换开销）

void doRespond(Call call) throws IOException {
      synchronized (call.connection.responseQueue) {
        call.connection.responseQueue.addLast(call);
        if (call.connection.responseQueue.size() == 1) {
          processResponse(call.connection.responseQueue, true);
        }
      }
    }

Response有一个类似于Listener的NIO选择器，用来处理当队列不为1时的发送。只是Listener关注OP_READ和OP_ACCEPT事件，而Response关注OP_WRITE事件。代码如下：

public void run() {

      while (running) {

          waitPending();     // 等待通道登记
          writeSelector.select(PURGE_INTERVAL); // 等待通道可写
          Iterator<SelectionKey> iter = writeSelector.selectedKeys().iterator();
          while (iter.hasNext()) {
            SelectionKey key = iter.next();
            iter.remove();
            try {
              if (key.isValid() && key.isWritable()) {
                  doAsyncWrite(key);//输出远程IPC调用结果
              }
            } catch (IOException e) {
            }
          }
          ......
    }
}

private void doAsyncWrite(SelectionKey key) throws IOException {
      Call call = (Call)key.attachment();
      ......
      synchronized(call.connection.responseQueue) {
        if (processResponse(call.connection.responseQueue, false)) {//调用输出
          try {
            key.interestOps(0);//processResponse返回true，表示无等待数据，清楚兴趣操作集
          } catch (CancelledKeyException e) {
           ......
          }
        }
      }
}

private boolean processResponse(LinkedList<Call> responseQueue,
                                    boolean inHandler) throws IOException {
      ......
    synchronized (responseQueue) {
      ......
      int numBytes = channelWrite(channel, call.response);

      done = true;               // error. no more data for this channel.
      closeConnection(call.connection);
    }
  return done;
}

processResponse关键点：

1. 可被Handler调用（当应答队列为1），参数inHandler为true，也可被Response调用，参数inHandler为false，表示队列为1或更多。

2. 返回true，表示通道上无需要发送的数据。

2.3总结

IPC Client端，发送Client.Call(new Invocation(method,args), remoteId)

--封装过程：Call.Id ,  Invocation---(查看Client.Connection.sendParam)

IPC Server端，接收Server.Call(Id, Invocation, Connction)---封装过程：Call.Id，Invocation--(查看Server.Connction.processData)