本套系列博客从真实商业环境抽取案例进行总结和分享，并给出Spark源码解读及商业实战指导，请持续关注本套博客。版权声明：本套Spark源码解读及商业实战归作者（秦凯新）所有，禁止转载，欢迎学习。

Spark商业环境实战及调优进阶系列

1. Spark 内置框架rpc通讯机制

TransportContext 内部握有创建TransPortClient和TransPortServer的方法实现，但却属于最底层的RPC通讯设施。为什么呢？

因为成员变量RPCHandler是抽象的，并没有具体的消息处理，而且TransportContext功能也在于创建TransPortClient客户端和TransPortServer服务端。具体解释如下：

 Contains the context to create a {@link TransportServer}, {@link TransportClientFactory}, and to
 setup Netty Channel pipelines with a
 {@link org.apache.spark.network.server.TransportChannelHandler}.

所以TransportContext只能为最底层的通讯基础。上层为NettyRPCEnv高层封装，并持有TransportContext引用，在TransportContext中传入NettyRpcHandler实体，来实现netty通讯回调Handler处理。TransportContext代码片段如下：

 /* The TransportServer and TransportClientFactory both create a TransportChannelHandler for each
 * channel. As each TransportChannelHandler contains a TransportClient, this enables server
 * processes to send messages back to the client on an existing channel.
 */
  public class TransportContext {
  private final Logger logger = LoggerFactory.getLogger(TransportContext.class);
  private final TransportConf conf;
  private final RpcHandler rpcHandler;
  private final boolean closeIdleConnections;

  private final MessageEncoder encoder;
  private final MessageDecoder decoder;

  public TransportContext(TransportConf conf, RpcHandler rpcHandler) {
    this(conf, rpcHandler, false);
  }

1.1 客户端和服务端统一的消息接收处理器 TransportChannelHandlerer

TransportClient 和TransportServer 在配置Netty的pipeLine的handler处理器时，均采用TransportChannelHandler, 来做统一的消息receive处理。为什么呢？在于统一消息处理入口，TransportChannelHandlerer根据消息类型执行不同的处理，代码片段如下：

 public void channelRead(ChannelHandlerContext ctx, Object request) throws Exception {
    if (request instanceof RequestMessage) {
      requestHandler.handle((RequestMessage) request);
   } else if (request instanceof ResponseMessage) {
      responseHandler.handle((ResponseMessage) request);
   } else {
      ctx.fireChannelRead(request);
   }

}

TransportContext初始化Pipeline的代码片段：

  public TransportChannelHandler initializePipeline(
  SocketChannel channel,
  RpcHandler channelRpcHandler) {
  try {
    
  TransportChannelHandler channelHandler = createChannelHandler(channel,
  
  channelRpcHandler);
  channel.pipeline()
    .addLast("encoder", ENCODER)
    .addLast(TransportFrameDecoder.HANDLER_NAME, NettyUtils.createFrameDecoder())
    .addLast("decoder", DECODER)
    .addLast("idleStateHandler", new IdleStateHandler(0, 0,   
                   conf.connectionTimeoutMs() / 1000))
                   
    .addLast("handler", channelHandler);
    
  return channelHandler;
} catch (RuntimeException e) {
  logger.error("Error while initializing Netty pipeline", e);
  throw e;
}

客户端和服务端统一的消息接收处理器 TransportChannelHandlerer 是这个函数:createChannelHandler(channel, channelRpcHandler)实现的，也即统一了这个netty的消息接受处理，代码片段如下：

    /**
    * Creates the server- and client-side handler which is used to handle both RequestMessages and
    * ResponseMessages. The channel is expected to have been successfully created, though certain
    * properties (such as the remoteAddress()) may not be available yet.
    */
    
    private TransportChannelHandler createChannelHandler(Channel channel,                                    RpcHandler rpcHandler) {
    
    TransportResponseHandler responseHandler = new                     
    TransportResponseHandler(channel);
    TransportClient client = new TransportClient(channel, responseHandler);
    
    TransportRequestHandler requestHandler = new TransportRequestHandler(channel, client,
    rpcHandler, conf.maxChunksBeingTransferred());
    
    return new TransportChannelHandler(client, responseHandler, requestHandler,
        conf.connectionTimeoutMs(), closeIdleConnections);
    }

不过transportClient对应的是TransportResponseHander，TransportServer对应的的是TransportRequestHander。在进行消息处理时，首先会经过TransportChannelHandler根据消息类型进行处理器选择，分别进行netty的消息生命周期管理：

exceptionCaught
channelActive
channelInactive
channelRead
userEventTriggered

1.2 transportClient对应的是ResponseMessage

客户端一旦发送消息（均为Request消息），就会在

private final Map<Long, RpcResponseCallback> outstandingRpcs;

private final Map<StreamChunkId, ChunkReceivedCallback> outstandingFetches

中缓存，用于回调处理。

1.3 transportServer对应的是RequestMessage

服务端接收消息类型（均为Request消息）

ChunkFetchRequest
RpcRequest
OneWayMessage
StremRequest

服务端响应类型（均为Response消息）：

ChunkFetchSucess
ChunkFetchFailure
RpcResponse
RpcFailure

2. Spark RpcEnv基础设施

2.1 上层建筑NettyRPCEnv

上层建筑NettyRPCEnv，持有TransportContext引用，在TransportContext中传入NettyRpcHandler实体，来实现netty通讯回调Handler处理

Dispatcher
TransportContext
TransPortClientFactroy
TransportServer
TransportConf

2.2 RpcEndPoint 与 RPCEndPointRef 端点

RpcEndPoint 为服务端
RPCEndPointRef 为客户端

2.2 Dispacher 与 Inbox 与 Outbox

一个端点对应一个Dispacher，一个Inbox , 多个OutBox

RpcEndpoint：RPC端点，Spark针对于每个节点（Client/Master/Worker）都称之一个Rpc端点 ,且都实现RpcEndpoint接口，内部根据不同端点的需求，设计不同的消息和不同的业务处理，如果需要发送（询问）则调用Dispatcher
RpcEnv：RPC上下文环境，每个Rpc端点运行时依赖的上下文环境称之为RpcEnv
Dispatcher：消息分发器，针对于RPC端点需要发送消息或者从远程RPC接收到的消息，分发至对应的指令收件箱/发件箱。如果指令接收方是自己存入收件箱，如果指令接收方为非自身端点，则放入发件箱
Inbox：指令消息收件箱，一个本地端点对应一个收件箱，Dispatcher在每次向Inbox存入消息时，都将对应EndpointData加入内部待Receiver Queue中，另外Dispatcher创建时会启动一个单独线程进行轮询Receiver Queue，进行收件箱消息消费
OutBox：指令消息发件箱，一个远程端点对应一个发件箱，当消息放入Outbox后，紧接着将消息通过TransportClient发送出去。消息放入发件箱以及发送过程是在同一个线程中进行，这样做的主要原因是远程消息分为RpcOutboxMessage, OneWayOutboxMessage两种消息，而针对于需要应答的消息直接发送且需要得到结果进行处理
TransportClient：Netty通信客户端，根据OutBox消息的receiver信息，请求对应远程TransportServer
TransportServer：Netty通信服务端，一个RPC端点一个TransportServer,接受远程消息后调用Dispatcher分发消息至对应收发件箱

Spark在Endpoint的设计上核心设计即为Inbox与Outbox，其中Inbox核心要点为：

内部的处理流程拆分为多个消息指令（InboxMessage）存放入Inbox
当Dispatcher启动最后，会启动一个名为【dispatcher-event-loop】的线程扫描Inbox待处理InboxMessage，并调用Endpoint根据InboxMessage类型做相应处理
当Dispatcher启动最后，默认会向Inbox存入OnStart类型的InboxMessage，Endpoint在根据OnStart指令做相关的额外启动工作，端点启动后所有的工作都是对OnStart指令处理衍生出来的，因此可以说OnStart指令是相互通信的源头。

注意：一个端点对应一个Dispacher，一个Inbox , 多个OutBox，可以看到 inbox在Dispacher 中且在EndPointData内部:

 private final RpcHandler rpcHandler;
/**
* A message dispatcher, responsible for routing RPC messages to the appropriate endpoint(s).
*/
 private[netty] class Dispatcher(nettyEnv: NettyRpcEnv) extends Logging {
 private class EndpointData(
    val name: String,
    val endpoint: RpcEndpoint,
    val ref: NettyRpcEndpointRef) {
  val inbox = new Inbox(ref, endpoint)
}
private val endpoints = new ConcurrentHashMap[String, EndpointData]
private val endpointRefs = new ConcurrentHashMap[RpcEndpoint, RpcEndpointRef]

// Track the receivers whose inboxes may contain messages.
private val receivers = new LinkedBlockingQueue[EndpointData]

注意：一个端点对应一个Dispacher，一个Inbox , 多个OutBox，可以看到 OutBox在NettyRpcEnv内部：

private[netty] class NettyRpcEnv(
  val conf: SparkConf,
  javaSerializerInstance: JavaSerializerInstance,
  host: String,
  securityManager: SecurityManager) extends RpcEnv(conf) with Logging {
  
  private val dispatcher: Dispatcher = new Dispatcher(this)
  
  private val streamManager = new NettyStreamManager(this)
  private val transportContext = new TransportContext(transportConf,
  new NettyRpcHandler(dispatcher, this, streamManager))
  
/**
 * A map for [[RpcAddress]] and [[Outbox]]. When we are connecting to a remote [[RpcAddress]],
 * we just put messages to its [[Outbox]] to implement a non-blocking `send` method.
 */
private val outboxes = new ConcurrentHashMap[RpcAddress, Outbox]()

2.3 Dispacher 与 Inbox 与 Outbox

Dispatcher的代码片段中，包含了核心的消息发送代码逻辑，意思是：向服务端发送一条消息，也即同时放进Dispatcher中的receiverrs中，也放进inbox的messages中。这个高层封装，如Master和Worker端点发送消息都是通过NettyRpcEnv中的 Dispatcher来实现的。在Dispatcher中有一个线程，叫做MessageLoop，实现消息的及时处理。

 /**
 * Posts a message to a specific endpoint.
 *
 * @param endpointName name of the endpoint.
 * @param message the message to post
  * @param callbackIfStopped callback function if the endpoint is stopped.
 */
 private def postMessage(
  endpointName: String,
  message: InboxMessage,
  callbackIfStopped: (Exception) => Unit): Unit = {
   val error = synchronized {
   val data = endpoints.get(endpointName)
   
  if (stopped) {
    Some(new RpcEnvStoppedException())
  } else if (data == null) {
    Some(new SparkException(s"Could not find $endpointName."))
  } else {
  
    data.inbox.post(message)
    receivers.offer(data)
    
    None
  }
 }

注意：默认第一条消息为onstart,为什么呢？看这里：

看到下面的 new EndpointData(name, endpoint, endpointRef) 了吗？

def registerRpcEndpoint(name: String, endpoint: RpcEndpoint): NettyRpcEndpointRef = {
 val addr = RpcEndpointAddress(nettyEnv.address, name)
    val endpointRef = new NettyRpcEndpointRef(nettyEnv.conf, addr, nettyEnv)
    synchronized {
  if (stopped) {
    throw new IllegalStateException("RpcEnv has been stopped")
  }
  if (endpoints.putIfAbsent(name, new EndpointData(name, endpoint, endpointRef)) != null) {
    throw new IllegalArgumentException(s"There is already an RpcEndpoint called $name")
  }
  val data = endpoints.get(name)
  endpointRefs.put(data.endpoint, data.ref)
  receivers.offer(data)  // for the OnStart message
}
endpointRef

}

注意EndpointData里面包含了inbox，因此Inbox初始化的时候，放进了onstart

 private class EndpointData(
  val name: String,
  val endpoint: RpcEndpoint,
  val ref: NettyRpcEndpointRef) {
val inbox = new Inbox(ref, endpoint)

}

onstart在Inbox初始化时出现了，注意每一个端点只有一个inbox，比如：master 节点。

2.4 发送消息流程为分为两种，一种端点(Master)自己把消息发送到本地Inbox，一种端点(Master)接收到消息后，通过TransPortRequestHander接收后处理，扔进Inbox

2.4.1 端点(Master)自己把消息发送到本地Inbox

- endpoint(Master) -> NettyRpcEnv-> Dispatcher ->  postMessage -> MessageLoop(Dispatcher) -> inbox -> process -> endpoint.receiveAndReply

解释如下：端点通过自己的RPCEnv环境，向自己的Inbox中发送消息，然后交由Dispatch来进行消息的处理，调用了端点自己的receiveAndReply方法

这里着重讲一下MessageLoop是什么时候启动的，参照Dispatcher的代码段如下，一旦初始化就会启动,因为是成员变量:

  private val threadpool: ThreadPoolExecutor = {
  val numThreads = nettyEnv.conf.getInt("spark.rpc.netty.dispatcher.numThreads",
    math.max(2, Runtime.getRuntime.availableProcessors()))
  val pool = ThreadUtils.newDaemonFixedThreadPool(numThreads, "dispatcher-event-loop")
  for (i <- 0 until numThreads) {
    pool.execute(new MessageLoop)
  }
   pool
 }

接着讲nettyRpcEnv是何时初始化的，Dispatcher是何时初始化的？

master初始化RpcEnv环境时，调用NettyRpcEnvFactory().create(config)进行初始化nettyRpcEnv，然后其成员变量Dispatcher开始初始化，然后Dispatcher内部成员变量threadpool开始启动messageLoop,然后开始处理消息，可谓是一环套一环啊。如下是Master端点初始化RPCEnv。

在NettyRpcEnv中，NettyRpcEnvFactory的create方法如下：

其中nettyRpcEnv.startServer，代码段如下，然后调用底层 transportContext.createServer来创建Server,并初始化netty 的 pipeline：

    server = transportContext.createServer(host, port, bootstraps)
    dispatcher.registerRpcEndpoint(
     RpcEndpointVerifier.NAME, new RpcEndpointVerifier(this, dispatcher))

最终端点开始不断向自己的Inboxz中发送消息即可，代码段如下：

    private def postMessage(
      endpointName: String,
      message: InboxMessage,
      callbackIfStopped: (Exception) => Unit): Unit = {
      error = synchronized {
      val data = endpoints.get(endpointName)
      if (stopped) {
           Some(new RpcEnvStoppedException())
      } else if (data == null) {
          Some(new SparkException(s"Could not find $endpointName."))
      } else {
      
         data.inbox.post(message)
         receivers.offer(data)
         
         None
      }
    }

2.4.2 端点(Master)接收到消息后，通过TransPortRequestHander接收后处理，扔进Inbox

- endpointRef(Worker) ->TransportChannelHandler -> channelRead0 -> TransPortRequestHander -> handle -> processRpcRequest ->NettyRpcHandler（在NettyRpcEnv中）  -> receive ->  internalReceive -> dispatcher.postToAll(RemoteProcessConnected(remoteEnvAddress)) （响应）-> dispatcher.postRemoteMessage(messageToDispatch, callback) （发送远端来的消息放进inbox）-> postMessage -> inbox -> process

如下图展示了整个消息接收到inbox的流程：

下图展示了 TransportChannelHandler接收消息:

    @Override
 public void channelRead0(ChannelHandlerContext ctx, Message request) throws Exception {
 if (request instanceof RequestMessage) {
  requestHandler.handle((RequestMessage) request);
} else {
  responseHandler.handle((ResponseMessage) request);
}
 }

然后TransPortRequestHander来进行消息匹配处理：

最终交给inbox的process方法，实际上由端点 endpoint.receiveAndReply(context)方法处理：

 /**
 * Process stored messages.
 */
 def process(dispatcher: Dispatcher): Unit = {
  var message: InboxMessage = null
    inbox.synchronized {
  if (!enableConcurrent && numActiveThreads != 0) {
    return
  }
  message = messages.poll()
  if (message != null) {
    numActiveThreads += 1
  } else {
    return
  }
}
while (true) {
  safelyCall(endpoint) {
    message match {
      case RpcMessage(_sender, content, context) =>
        try {
          endpoint.receiveAndReply(context).applyOrElse[Any, Unit](content, { msg =>
            throw new SparkException(s"Unsupported message $message from ${_sender}")
          })
        } catch {
          case NonFatal(e) =>
            context.sendFailure(e)
            // Throw the exception -- this exception will be caught by the safelyCall function.
            // The endpoint's onError function will be called.
            throw e
        }

      case OneWayMessage(_sender, content) =>
        endpoint.receive.applyOrElse[Any, Unit](content, { msg =>
          throw new SparkException(s"Unsupported message $message from ${_sender}")
        })

      case OnStart =>
        endpoint.onStart()
        if (!endpoint.isInstanceOf[ThreadSafeRpcEndpoint]) {
          inbox.synchronized {
            if (!stopped) {
              enableConcurrent = true
            }
          }
        }

      case OnStop =>
        val activeThreads = inbox.synchronized { inbox.numActiveThreads }
        assert(activeThreads == 1,
          s"There should be only a single active thread but found $activeThreads threads.")
        dispatcher.removeRpcEndpointRef(endpoint)
        endpoint.onStop()
        assert(isEmpty, "OnStop should be the last message")

      case RemoteProcessConnected(remoteAddress) =>
        endpoint.onConnected(remoteAddress)

      case RemoteProcessDisconnected(remoteAddress) =>
        endpoint.onDisconnected(remoteAddress)

      case RemoteProcessConnectionError(cause, remoteAddress) =>
        endpoint.onNetworkError(cause, remoteAddress)
    }
  }

  inbox.synchronized {
    // "enableConcurrent" will be set to false after `onStop` is called, so we should check it
    // every time.
    if (!enableConcurrent && numActiveThreads != 1) {
      // If we are not the only one worker, exit
      numActiveThreads -= 1
      return
    }
    message = messages.poll()
    if (message == null) {
      numActiveThreads -= 1
      return
    }
  }
}

}

3 结语

本文花了将近两天时间进行剖析Spark的 Rpc 工作原理，真是不容易，关键是你看懂了吗？欢迎评论

秦凯新