Spark2.2——RpcEnv(一)

版权声明:本文为博主原创文章,未经博主允许不得转载。 https://blog.csdn.net/Jorocco/article/details/86490480

  Spark1.6推出的RpcEnv、RpcEndPoint、RpcEndpointRef为核心的新型架构下的RPC通信方式,在底层封装了Akka和Netty,为未来扩充更多的通信系统提供了可能。RpcEnv是一个更宏观的Env,是Spark集群Rpc通信的基础服务环境,因此在集群启动时候所有的节点(无论Master还是Worker)都会创建一个RpcEnv,然后将该节点注册到RpcEnv中。RpcEnv是RPC的环境,所有的RpcEndpoint都需要注册到RpcEnv实例对象中,管理着这些注册的RpcEndpoint的生命周期:

  • 根据name或者uri注册RpcEndpoint;
  • 管理各种消息的处理;
  • 停止RpcEndpoint

  Spark RPC中最为重要的三个抽象(“三剑客”)为:RpcEnv、RpcEndpoint、RpcEndpointRef,这样做的好处有:

  • 对上层的API来说,屏蔽了底层的具体实现,使用方便
  • 可以通过不同的实现来完成指定的功能,方便扩展
  • 促进了底层实现层的良性竞争,Spark 1.6.3中默认使用了Netty作为底层的实现,但Akka的依赖依然存在;而Spark2.2.0中的底层实现只有Netty,这样用户可以方便的使用不同版本的Akka或者将来某种更好的底层实现

创建RpcEnv的代码:

/*
	变量声明
	包名:org.apache.spark
	类名:SparkEnv
*/
private[spark] val driverSystemName = "sparkDriver"
private[spark] val executorSystemName = "sparkExecutor"
  
val isDriver = executorId == SparkContext.DRIVER_IDENTIFIER
  
val systemName = if (isDriver) driverSystemName else executorSystemName
val rpcEnv = RpcEnv.create(systemName, bindAddress, advertiseAddress, port.getOrElse(-1), conf,
  securityManager, numUsableCores, !isDriver)
/*
	变量处理
	第一步
	包名:org.apache.spark.rpc
	类名:RpcEnv
*/
def create(
    name: String,
    bindAddress: String,
    advertiseAddress: String,
    port: Int,
    conf: SparkConf,
    securityManager: SecurityManager,
    numUsableCores: Int,
    clientMode: Boolean): RpcEnv = {
  val config = RpcEnvConfig(conf, name, bindAddress, advertiseAddress, port, securityManager,
    numUsableCores, clientMode)
  new NettyRpcEnvFactory().create(config)
}

/*
	第二步
	包名:org.apache.spark.rpc.netty
	类名:NettyRpcEnv
*/
private[rpc] class NettyRpcEnvFactory extends RpcEnvFactory with Logging {

  /**
    * NettyRpcEnvFactory创建了NettyRpcEnv之后,如果clientMode为false,即服务端(Driver端Rpc通讯),则使用创建出
    * 的NettyRpcEnv的函数startServer定义一个函数变量startNettyRpcEnv((nettyEnv, nettyEnv.address.port)为函
    * 数的返回值),将该函数作为参数传递给函数Utils.startServiceOnPort,即在Driver端启动服务。
    *
    * 这里可以进入Utils.startServiceOnPort这个函数看看源代码,可以看出为什么不直接调用nettyEnv.startServer,而要把它封装起来
    * 传递给工具类来调用:在这个端口启动服务不一定一次就能成功,工具类里对失败的情况做最大次数的尝试,直到启动成功并返回启
    * 动成功后的端口。
    */
  def create(config: RpcEnvConfig): RpcEnv = {
    val sparkConf = config.conf
    // Use JavaSerializerInstance in multiple threads is safe. However, if we plan to support
    // KryoSerializer in future, we have to use ThreadLocal to store SerializerInstance
    // 在多个线程中使用JavaSerializerInstance是安全的。然而,如果我们计划将来支持KryoSerializer,
    // 我们必须使用ThreadLocal来存储SerializerInstance
    // Netty的通讯都是基于Jav序列化,暂时不支持Kryo

    // 1.初始化JavaSerializer,初始化NettyRpcEnv,如果是 非客户端模式就启动netty服务
    val javaSerializerInstance =
      new JavaSerializer(sparkConf).newInstance().asInstanceOf[JavaSerializerInstance]

    // 2.初始化NettyRpcEnv
    val nettyEnv =
      new NettyRpcEnv(sparkConf, javaSerializerInstance, config.advertiseAddress,
        config.securityManager)
    // 3.判断是否在Driver端,如果在,则构建TransportServer并注册dispatcher,否则直接返回nettyEnv 
    if (!config.clientMode) {
      // startNettyRpcEnv作为一个函数变量将在下面的startServiceOnPort中被调用
      // 简单解释一下这段代码
      // 声明一个函数变量,参数是int(actuslPort),=>后面是实现体,最终返回的是2元组(NettyRpcEnv,int)
      val startNettyRpcEnv: Int => (NettyRpcEnv, Int) = { actualPort =>
        /** 主要是构建TransportServer和注册dispatcher */
        nettyEnv.startServer(config.bindAddress, actualPort)
        (nettyEnv, nettyEnv.address.port)
      }
      try {
        // 其实内部实现还是调用startNettyRpcEnv在指定的端口实例化,并且返回nettyEnv对象
        Utils.startServiceOnPort(config.port, startNettyRpcEnv, sparkConf, config.name)._1
      } catch {
        case NonFatal(e) =>
          nettyEnv.shutdown()
          throw e
      }
    }
    nettyEnv
  }
}

/*
	包名:org.apache.spark.rpc.netty
	类名:NettyRpcEnv
*/
def startServer(bindAddress: String, port: Int): Unit = {
    val bootstraps: java.util.List[TransportServerBootstrap] =
      // 检查是否启用了Spark通信协议的身份验证。
      if (securityManager.isAuthenticationEnabled()) {
        // Spark的auth协议进行身份验证
        java.util.Arrays.asList(new AuthServerBootstrap(transportConf, securityManager))
      } else {
        java.util.Collections.emptyList()
      }
    // 创建TransportServer
    server = transportContext.createServer(bindAddress, port, bootstraps)
    // 创建RpcEndpointVerifier,然后注册自己到NettyRpcEnv上并发回自己的Ref的实现
    dispatcher.registerRpcEndpoint(
      RpcEndpointVerifier.NAME, new RpcEndpointVerifier(this, dispatcher))
  }

 /**
  包名:org.apache.spark.util
  类名:Utils
  /**
  启动的时候master at spark://biluos.com:7079 后面的端口号开始从7077开始 一直到成功
   */
  def startServiceOnPort[T](
      startPort: Int,
      startService: Int => (T, Int),
      conf: SparkConf,
      serviceName: String = ""): (T, Int) = {

    /**端口号必须1024 and 65535 之间*/
    require(startPort == 0 || (1024 <= startPort && startPort < 65536),
      "startPort should be between 1024 and 65535 (inclusive), or 0 for a random free port.")

    val serviceString = if (serviceName.isEmpty) "" else s" '$serviceName'"
    val maxRetries = portMaxRetries(conf)
    for (offset <- 0 to maxRetries) {
      // Do not increment port if startPort is 0, which is treated as a special port
      val tryPort = if (startPort == 0) {
        startPort
      } else {
        userPort(startPort, offset)
      }
      try {
        val (service, port) = startService(tryPort)
        // 17/12/05 11:56:50 INFO Utils: Successfully started service 'sparkDriver' on port 55271.
        //22=> 17/12/05 11:56:50 INFO Utils: Successfully started service 'SparkUI' on port 4040.
        //25=>17/12/05 11:56:51 INFO Utils: Successfully started service 'org.apache.spark.network.netty.NettyBlockTransferService' on port 55290.
        logInfo(s"Successfully started service$serviceString on port $port.")
        return (service, port)
      } catch {
        case e: Exception if isBindCollision(e) =>
          if (offset >= maxRetries) {
            val exceptionMessage = if (startPort == 0) {
              s"${e.getMessage}: Service$serviceString failed after " +
                s"$maxRetries retries (on a random free port)! " +
                s"Consider explicitly setting the appropriate binding address for " +
                s"the service$serviceString (for example spark.driver.bindAddress " +
                s"for SparkDriver) to the correct binding address."
            } else {
              s"${e.getMessage}: Service$serviceString failed after " +
                s"$maxRetries retries (starting from $startPort)! Consider explicitly setting " +
                s"the appropriate port for the service$serviceString (for example spark.ui.port " +
                s"for SparkUI) to an available port or increasing spark.port.maxRetries."
            }
            val exception = new BindException(exceptionMessage)
            // restore original stack trace
            exception.setStackTrace(e.getStackTrace)
            throw exception
          }
          if (startPort == 0) {
            // As startPort 0 is for a random free port, it is most possibly binding address is
            // not correct.
            logWarning(s"Service$serviceString could not bind on a random free port. " +
              "You may check whether configuring an appropriate binding address.")
          } else {
            logWarning(s"Service$serviceString could not bind on port $tryPort. " +
              s"Attempting port ${tryPort + 1}.")
          }
      }
    }
    // Should never happen
    throw new SparkException(s"Failed to start service$serviceString on port $startPort")
  }

RpcEnv、RpcEndPoint、RpcEndpointRef更深入的剖析详见《Spark2.2——RpcEnv(二)》

展开阅读全文

没有更多推荐了,返回首页