Akka（25）： Stream：对接外部系统-Integration

Posted 2020-10-07 雪川大虫

   在现实应用中akka-stream往往需要集成其它的外部系统形成完整的应用。这些外部系统可能是akka系列系统或者其它类型的系统。所以，akka-stream必须提供一些函数和方法来实现与各种不同类型系统的信息交换。在这篇讨论里我们就介绍几种通用的信息交换方法和函数。

   akka-stream提供了mapAsync+ask模式可以从一个运算中的数据流向外连接某个Actor来进行数据交换。这是一种akka-stream与Actor集成的应用。说到与Actor集成，联想到如果能把akka-stream中复杂又消耗资源的运算任务交付给Actor，那么我们就可以充分利用actor模式的routing,cluster,supervison等等特殊功能来实现分布式高效安全的运算。下面就是这个mapAsync函数定义：

  /**
   * Transform this stream by applying the given function to each of the elements
   * as they pass through this processing step. The function returns a `Future` and the
   * value of that future will be emitted downstream. The number of Futures
   * that shall run in parallel is given as the first argument to ``mapAsync``.
   * These Futures may complete in any order, but the elements that
   * are emitted downstream are in the same order as received from upstream.
   *
   * If the function `f` throws an exception or if the `Future` is completed
   * with failure and the supervision decision is [[akka.stream.Supervision.Stop]]
   * the stream will be completed with failure.
   *
   * If the function `f` throws an exception or if the `Future` is completed
   * with failure and the supervision decision is [[akka.stream.Supervision.Resume]] or
   * [[akka.stream.Supervision.Restart]] the element is dropped and the stream continues.
   *
   * The function `f` is always invoked on the elements in the order they arrive.
   *
   * Adheres to the [[ActorAttributes.SupervisionStrategy]] attribute.
   *
   * ‘‘‘Emits when‘‘‘ the Future returned by the provided function finishes for the next element in sequence
   *
   * ‘‘‘Backpressures when‘‘‘ the number of futures reaches the configured parallelism and the downstream
   * backpressures or the first future is not completed
   *
   * ‘‘‘Completes when‘‘‘ upstream completes and all futures have been completed and all elements have been emitted
   *
   * ‘‘‘Cancels when‘‘‘ downstream cancels
   *
   * @see [[#mapAsyncUnordered]]
   */
  def mapAsync[T](parallelism: Int)(f: Out ? Future[T]): Repr[T] = via(MapAsync(parallelism, f))

mapAsync把一个函数f: Out=>Future[T]在parallelism个Future里并行运算。我们来看看ask的款式：

  def ?(message: Any)(implicit timeout: Timeout, sender: ActorRef = Actor.noSender): Future[Any] =
    internalAsk(message, timeout, sender)

刚好是 T=>Future[T]这样的款式。所以我们可以用下面这种方式从Stream里与Actor沟通：

  stream.mapAsync(parallelism = 5)(elem => (ref ? elem).mapTo[String])

在以上的用例里Stream的每一个元素都通过ref ? elem发送给了ActorRef在一个Future里运算，这个Actor完成运算后返回Future[String]类型结果。值得注意的是mapAsync通过这个返回的Future来实现stream backpressure，也就是说这个运算Actor必须返回结果，否则Stream就会挂在那里了。下面我们先示范一下mapAsync的直接应用：

import akka.actor._
import akka.pattern._
import akka.stream._
import akka.stream.scaladsl._
import akka.routing._

import scala.concurrent.duration._
import akka.util.Timeout

object StorageActor {

  case class Query(rec: Int, qry: String) //模拟数据存写Query

  class StorageActor extends Actor with ActorLogging { //模拟存写操作Actor
    override def receive: Receive = {
      case Query(num,qry) =>
        val reply = s"${self.path} is saving: [$qry]"
        sender() ! reply                  //必须回复mapAsync, 抵消backpressure
        reply
    }
  }
  def props = Props(new StorageActor)
}

object MapAsyncDemo extends App {
  implicit val sys = ActorSystem("demoSys")
  implicit val ec = sys.dispatcher
  implicit val mat = ActorMaterializer(
    ActorMaterializerSettings(sys)
      .withInputBuffer(initialSize = 16, maxSize = 16)
  )
  val storageActor = sys.actorOf(StorageActor.props,"dbWriter")


  implicit val timeout = Timeout(3 seconds)
  Source(Stream.from(1)).delay(1.second,DelayOverflowStrategy.backpressure)
    .mapAsync(parallelism = 3){ n =>
      (storageActor ? StorageActor.Query(n,s"Record#$n")).mapTo[String]
    }.runWith(Sink.foreach(println))

  scala.io.StdIn.readLine()
  sys.terminate()

}

 在这个例子里parallelism=3，我们在StorageActor里把当前运算中的实例返回并显示出来：

akka://demoSys/user/dbWriter is saving: [Record#1]
akka://demoSys/user/dbWriter is saving: [Record#2]
akka://demoSys/user/dbWriter is saving: [Record#3]
akka://demoSys/user/dbWriter is saving: [Record#4]
akka://demoSys/user/dbWriter is saving: [Record#5]
akka://demoSys/user/dbWriter is saving: [Record#6]
...

可以看到：mapAsync只调用了一个Actor。那么所谓的并行运算parallelism=3的意思就只能代表在多个Future线程中同时运算了。为了实现对Actor模式特点的充分利用，我们可以通过router来实现在多个actor上并行运算。Router分pool和group两种类型：pool类router自己构建routees，group类型则调用已经构建的Actor。在我们这次的测试里只能使用group类型的Router，因为如果需要对routee实现监管supervision的话，pool类型的router在routee终止时会自动补充构建新的routee，如此就避开了监管策略。首先增加StorageActor的routing功能：

  val numOfActors = 3
  val routees: List[ActorRef] = List.fill(numOfActors)(      //构建3个StorageActor
    sys.actorOf(StorageActor.props))
  val routeePaths: List[String] = routees.map{ref => "/user/"+ref.path.name}

  val storageActorPool = sys.actorOf(
    RoundRobinGroup(routeePaths).props()
      .withDispatcher("akka.pool-dispatcher")
    ,"starageActorPool"
  )

  implicit val timeout = Timeout(3 seconds)
  Source(Stream.from(1)).delay(1.second,DelayOverflowStrategy.backpressure)
    .mapAsync(parallelism = 1){ n =>
      (storageActorPool ? StorageActor.Query(n,s"Record#$n")).mapTo[String]
    }.runWith(Sink.foreach(println))

我们使用了RoundRobinGroup作为智能任务分配方式。注意上面的parallelism=1：现在不需要多个Future了。再看看运行的结果显示：

akka://demoSys/user/$a is saving: [Record#1]
akka://demoSys/user/$b is saving: [Record#2]
akka://demoSys/user/$c is saving: [Record#3]
akka://demoSys/user/$a is saving: [Record#4]
akka://demoSys/user/$b is saving: [Record#5]
akka://demoSys/user/$c is saving: [Record#6]
akka://demoSys/user/$a is saving: [Record#7]

可以看到现在运算任务是在a,b,c三个Actor上并行运算的。既然是模拟数据库的并行存写动作，我们可以试着为每个routee增加逐步延时重启策略BackOffSupervisor：

object StorageActor {

  case class Query(rec: Int, qry: String) //模拟数据存写Query
  class DbException(cause: String) extends Exception(cause) //自定义存写异常

  class StorageActor extends Actor with ActorLogging { //存写操作Actor
    override def receive: Receive = {
      case Query(num,qry) =>
        var res: String = ""
        try {
          res = saveToDB(num,qry)
        } catch {
          case e: Exception => Error(num,qry) //模拟操作异常
        }
        sender() ! res
      case _ =>
    }
    def saveToDB(num: Int,qry: String): String = { //模拟存写函数
      val msg = s"${self.path} is saving: [$qry#$num]"
      if ( num % 3 == 0) Error(num,qry)        //模拟异常
      else {
        log.info(s"${self.path} is saving: [$qry#$num]")
        s"${self.path} is saving: [$qry#$num]"
      }
    }
    def Error(num: Int,qry: String): String = {
      val msg = s"${self.path} is saving: [$qry#$num]"
      sender() ! msg
      throw new DbException(s"$msg blew up, boooooom!!!")
    }

    //验证异常重启
    //BackoffStrategy.onStop goes through restart process
    override def preRestart(reason: Throwable, message: Option[Any]): Unit = {
      log.info(s"Restarting ${self.path.name} on ${reason.getMessage}")
      super.preRestart(reason, message)
    }

    override def postRestart(reason: Throwable): Unit = {
      log.info(s"Restarted ${self.path.name} on ${reason.getMessage}")
      super.postRestart(reason)
    }

    override def postStop(): Unit = {
      log.info(s"Stopped ${self.path.name}!")
      super.postStop()
    }
   //BackOffStrategy.onFailure dosn‘t go through restart process
    override def preStart(): Unit = {
      log.info(s"PreStarting ${self.path.name} ...")
      super.preStart()
    }


  }
  def props = Props(new StorageActor)
}

object StorageActorGuardian {  //带监管策略的StorageActor
  def props: Props = { //在这里定义了监管策略和StorageActor构建
    def decider: PartialFunction[Throwable, SupervisorStrategy.Directive] = {
      case _: StorageActor.DbException => SupervisorStrategy.Restart
    }

    val options = Backoff.onStop(StorageActor.props, "dbWriter", 100 millis, 500 millis, 0.0)
      .withManualReset
      .withSupervisorStrategy(
        OneForOneStrategy(maxNrOfRetries = 3, withinTimeRange = 1 second)(
          decider.orElse(SupervisorStrategy.defaultDecider)
        )
      )
    BackoffSupervisor.props(options)
  }
}

object IntegrateDemo extends App {
  implicit val sys = ActorSystem("demoSys")
  implicit val ec = sys.dispatcher
  implicit val mat = ActorMaterializer(
    ActorMaterializerSettings(sys)
      .withInputBuffer(initialSize = 16, maxSize = 16)
  )

  val numOfActors = 3
  val routees: List[ActorRef] = List.fill(numOfActors)(
    sys.actorOf(StorageActorGuardian.props))
  val routeePaths: List[String] = routees.map{ref => "/user/"+ref.path.name} //获取ActorPath

  val storageActorPool = sys.actorOf(
    RoundRobinGroup(routeePaths).props()
      .withDispatcher("akka.pool-dispatcher")
    ,"starageActorPool"
  )

  implicit val timeout = Timeout(3 seconds)
  Source(Stream.from(1)).delay(3.second,DelayOverflowStrategy.backpressure)
    .mapAsync(parallelism = 1){ n =>
      (storageActorPool ? StorageActor.Query(n,s"Record")).mapTo[String]
    }.runWith(Sink.foreach(println))

  scala.io.StdIn.readLine()
  sys.terminate()

}

我们同时增加了模拟异常发生、StorageActor生命周期callback来跟踪异常发生时SupervisorStrategy.Restart的执行情况。从试运行反馈结果证实Backoff.onFailure不会促发Restart事件，而是直接促发了preStart事件。Backoff.onStop则走Restart过程。Backoff.onFailure是在Actor出现异常终止触动的，而Backoff.onStop则是目标Actor在任何情况下终止后触发。值得注意的是，在以上例子里运算Actor会越过造成异常的这个流元素，所以我们必须在preRestart里把这个元素补发给自己：

   //验证异常重启
    //BackoffStrategy.onStop goes through restart process
    override def preRestart(reason: Throwable, message: Option[Any]): Unit = {
      log.info(s"Restarting ${self.path.name} on ${reason.getMessage}")
       message match {
        case Some(Query(n,qry)) =>
          self ! Query(n+101,qry)      //把异常消息再补发送给自己，n+101更正了异常因素
        case _ =>
          log.info(s"Exception message: None")

      }
      super.preRestart(reason, message)
    }

如果我们不需要委托给Actor运算任务的返回结果，可以尝试用Sink.actorRefWithAck：

 /**
   * Sends the elements of the stream to the given `ActorRef` that sends back back-pressure signal.
   * First element is always `onInitMessage`, then stream is waiting for acknowledgement message
   * `ackMessage` from the given actor which means that it is ready to process
   * elements. It also requires `ackMessage` message after each stream element
   * to make backpressure work.
   *
   * If the target actor terminates the stream will be canceled.
   * When the stream is completed successfully the given `onCompleteMessage`
   * will be sent to the destination actor.
   * When the stream is completed with failure - result of `onFailureMessage(throwable)`
   * function will be sent to the destination actor.
   */
  def actorRefWithAck[T](ref: ActorRef, onInitMessage: Any, ackMessage: Any, onCompleteMessage: Any,
                         onFailureMessage: (Throwable) ? Any = Status.Failure): Sink[T, NotUsed] =
    Sink.fromGraph(new ActorRefBackpressureSinkStage(ref, onInitMessage, ackMessage, onCompleteMessage, onFailureMessage))

在这里ActorRef只能返回有关backpressure状态信号。actorRefWithAck自己则返回Sink[T,NotUsed]，也就是说它构建了一个Sink。actorRefWithAck使用三种信号来与目标Actor沟通：

1、onInitMessage：stream发送给ActorRef的第一个信号，表示可以开始数据交换

2、ackMessage：ActorRef向stream发出的信号，回复自身准备完毕，可以接收消息，也是一种backpressure卸除消息

3、onCompleteMessage：stream发给ActorRef，通知stream已经完成了所有流元素发送

我们必须修改上个例子中的StorageActor来符合actorRefWithAck的应用和与目标Actor的沟通：

object StorageActor {
  val onInitMessage = "start"
  val onCompleteMessage = "done"
  val ackMessage = "ack"

  case class Query(rec: Int, qry: String) //模拟数据存写Query
  class DbException(cause: String) extends Exception(cause) //自定义存写异常

  class StorageActor extends Actor with ActorLogging { //存写操作Actor
    override def receive: Receive = {
      case `onInitMessage` => sender() ! ackMessage
      case Query(num,qry) =>
        var res: String = ""
        try {
          res = saveToDB(num,qry)
        } catch {
          case e: Exception => Error(num,qry) //模拟操作异常
        }
        sender() ! ackMessage
      case `onCompleteMessage` => //clean up resources 释放资源
      case _ =>
    }
    def saveToDB(num: Int,qry: String): String = { //模拟存写函数
      val msg = s"${self.path} is saving: [$qry#$num]"
      if ( num % 5 == 0) Error(num,qry)        //模拟异常
      else {
        log.info(s"${self.path} is saving: [$qry#$num]")
        s"${self.path} is saving: [$qry#$num]"
      }
    }
    def Error(num: Int,qry: String) = {
      val msg = s"${self.path} is saving: [$qry#$num]"
      sender() ! ackMessage
      throw new DbException(s"$msg blew up, boooooom!!!")
    }

    //验证异常重启
    //BackoffStrategy.onStop goes through restart process
    override def preRestart(reason: Throwable, message: Option[Any]): Unit = {
      log.info(s"Restarting ${self.path.name} on ${reason.getMessage}")
      message match {
        case Some(Query(n,qry)) =>
          self ! Query(n+101,qry)      //把异常消息再补发送给自己，n+101更正了异常因素
        case _ =>
          log.info(s"Exception message: None")

      }
      super.preRestart(reason, message)
    }

    override def postRestart(reason: Throwable): Unit = {
      log.info(s"Restarted ${self.path.name} on ${reason.getMessage}")
      super.postRestart(reason)
    }

    override def postStop(): Unit = {
      log.info(s"Stopped ${self.path.name}!")
      super.postStop()
    }
    //BackOffStrategy.onFailure dosn‘t go through restart process
    override def preStart(): Unit = {
      log.info(s"PreStarting ${self.path.name} ...")
      super.preStart()
    }
  }
  def props = Props(new StorageActor)
}

StorageActor类里包括了对actorRefWithAck沟通消息onInitMessage、ackMessage、onCompleteMessage的处理。这个Actor只返回backpressure消息ackMessage，而不是返回任何运算结果。注意，在preRestart里我们把造成异常的元素处理后再补发给了自己。Sink.actorRefWithAck的调用方式如下： 

  Source(Stream.from(1)).map(n => Query(n,s"Record")).delay(3.second,DelayOverflowStrategy.backpressure)
      .runWith(Sink.actorRefWithAck(
       storageActorPool, onInitMessage, ackMessage,onCompleteMessage))

完整的运行环境源代码如下：

object SinkActorRefWithAck extends App {
  import StorageActor._

  implicit val sys = ActorSystem("demoSys")
  implicit val ec = sys.dispatcher
  implicit val mat = ActorMaterializer(
    ActorMaterializerSettings(sys)
      .withInputBuffer(initialSize = 16, maxSize = 16)
  )

  val storageActor = sys.actorOf(StorageActor.props,"storageActor")

  val numOfActors = 3
  val routees: List[ActorRef] = List.fill(numOfActors)(
    sys.actorOf(StorageActorGuardian.props))
  val routeePaths: List[String] = routees.map{ref => "/user/"+ref.path.name}

  val storageActorPool = sys.actorOf(
    RoundRobinGroup(routeePaths).props()
      .withDispatcher("akka.pool-dispatcher")
    ,"starageActorPool"
  )

  Source(Stream.from(1)).map(n => Query(n,s"Record")).delay(3.second,DelayOverflowStrategy.backpressure)
      .runWith(Sink.actorRefWithAck(
       storageActorPool, onInitMessage, ackMessage,onCompleteMessage))

  scala.io.StdIn.readLine()
  sys.terminate()

}

如果一个外部系统向一个数据流提供数据，那我们可以把这个外部系统当作数据流的源头Source。akka-stream提供了个Source.queque函数来构建一种Source，外部系统可以利用这个Source来向Stream发送数据。Source.queque的函数款式如下：

  /**
   * Creates a `Source` that is materialized as an [[akka.stream.scaladsl.SourceQueue]].
   * You can push elements to the queue and they will be emitted to the stream if there is demand from downstream,
   * otherwise they will be buffered until request for demand is received. Elements in the buffer will be discarded
   * if downstream is terminated.
   *
   * Depending on the defined [[akka.stream.OverflowStrategy]] it might drop elements if
   * there is no space available in the buffer.
   *
   * Acknowledgement mechanism is available.
   * [[akka.stream.scaladsl.SourceQueue.offer]] returns `Future[QueueOfferResult]` which completes with
   * `QueueOfferResult.Enqueued` if element was added to buffer or sent downstream. It completes with
   * `QueueOfferResult.Dropped` if element was dropped. Can also complete  with `QueueOfferResult.Failure` -
   * when stream failed or `QueueOfferResult.QueueClosed` when downstream is completed.
   *
   * The strategy [[akka.stream.OverflowStrategy.backpressure]] will not complete last `offer():Future`
   * call when buffer is full.
   *
   * You can watch accessibility of stream with [[akka.stream.scaladsl.SourceQueue.watchCompletion]].
   * It returns future that completes with success when stream is completed or fail when stream is failed.
   *
   * The buffer can be disabled by using `bufferSize` of 0 and then received message will wait
   * for downstream demand unless there is another message waiting for downstream demand, in that case
   * offer result will be completed according to the overflow strategy.
   *
   * @param bufferSize size of buffer in element count
   * @param overflowStrategy Strategy that is used when incoming elements cannot fit inside the buffer
   */
  def queue[T](bufferSize: Int, overflowStrategy: OverflowStrategy): Source[T, SourceQueueWithComplete[T]] =
    Source.fromGraph(new QueueSource(bufferSize, overflowStrategy).withAttributes(DefaultAttributes.queueSource))

Source.queue构建了一个Source：Source[T,SourceQueueWithComplete[T]]，SourceQueueWithComplete类型如下：

/**
 * This trait adds completion support to [[SourceQueue]].
 */
trait SourceQueueWithComplete[T] extends SourceQueue[T] {
  /**
   * Complete the stream normally. Use `watchCompletion` to be notified of this
   * operation’s success.
   */
  def complete(): Unit

  /**
   * Complete the stream with a failure. Use `watchCompletion` to be notified of this
   * operation’s success.
   */
  def fail(ex: Throwable): Unit

  /**
   * Method returns a [[Future]] that will be completed if the stream completes,
   * or will be failed when the stage faces an internal failure or the the [[SourceQueueWithComplete.fail]] method is invoked.
   */
  def watchCompletion(): Future[Done]
}

它在SourceQueue的基础上增加了几个抽象函数，主要用来向目标数据流发送终止信号包括：complete,fail。watchCompletion是一种监视函数，返回Future代表SourceQueue被手工终止或stream由于某些原因终止运算。下面是SourceQueue定义：

/**
 * This trait allows to have the queue as a data source for some stream.
 */
trait SourceQueue[T] {

  /**
   * Method offers next element to a stream and returns future that:
   * - completes with `Enqueued` if element is consumed by a stream
   * - completes with `Dropped` when stream dropped offered element
   * - completes with `QueueClosed` when stream is completed during future is active
   * - completes with `Failure(f)` when failure to enqueue element from upstream
   * - fails when stream is completed or you cannot call offer in this moment because of implementation rules
   * (like for backpressure mode and full buffer you need to wait for last offer call Future completion)
   *
   * @param elem element to send to a stream
   */
  def offer(elem: T): Future[QueueOfferResult]

  /**
   * Method returns a [[Future]] that will be completed if the stream completes,
   * or will be failed when the stage faces an internal failure.
   */
  def watchCompletion(): Future[Done]
}

这个界面支持了SourceQueue的基本操作：offer(elem: T), watchComplete()两个函数。下面我们就用个例子来示范SourceQueue的使用方法：我们用Calculator actor来模拟外部系统、先用Source.queue构建一个SourceQueue然后再连接下游形成一个完整的数据流。把这个数据流传给Calculator，这样Calculator就可以向这个运行中的Stream发送数据了。我们会通过这个过程来示范SourceQueue的基本操作。下面这个Calculator Actor模拟了一个外部系统作为SourceQueue用户：

object Calculator {
  trait Operations
  case class Add(op1:Int, op2:Int) extends Operations
  case class DisplayError(err: Exception) extends Operations
  case object Stop extends Operations
  case class ProduceError(err: Exception) extends Operations

  def props(inputQueue: SourceQueueWithComplete[String]) = Props(new Calculator(inputQueue))
}
class Calculator(inputQueue: SourceQueueWithComplete[String]) extends Actor with ActorLogging{
  import Calculator._
  import context.dispatcher
  override def receive: Receive = {
    case Add(op1,op2) =>
      val msg = s"$op1 + $op2 = ${op1 + op2}"
      inputQueue.offer(msg)    //.mapTo[String]
        .recover {
        case e: Exception => DisplayError(e)}
        .pipeTo(self)
    case QueueOfferResult.Enqueued =>
      log.info("QueueOfferResult.Enqueued")
    case QueueOfferResult.Dropped =>
    case QueueOfferResult.Failure(cause) =>
    case QueueOfferResult.QueueClosed  =>
      log.info("QueueOfferResult.QueueClosed")

    case Stop => inputQueue.complete()
    case ProduceError(e) => inputQueue.fail(e)

  }
}

我们看到，Calculator通过传入的inputQueue把计算结果传给数据流显示出来。在receive函数里我们把offer用法以及它可能产生的返回结果通过pipeTo都做了示范。注意：不能使用mapTo[String]，因为offer返回Future[T]，T不是String，会造成类型转换错误。而我们已经在Source.queue[String]注明了offer(elem) elem的类型是String。inputQueue的构建方式如下：

  val source: Source[String, SourceQueueWithComplete[String]]  =
               Source.queue[String](bufferSize = 16,
               overflowStrategy = OverflowStrategy.backpressure)

  val inputQueue: SourceQueueWithComplete[String] = source.toMat(Sink.foreach(println))(Keep.left).run()

  inputQueue.watchCompletion().onComplete {
    case Success(result) => println(s"Calculator ends with: $result")
    case Failure(cause)  => println(s"Calculator ends with exception: ${cause.getMessage}")
  }

增加了watchCompetion来监测SourceQueue发出的终止信号。我们还可以看到：以上SoureQueue实例source是支持backpressure的。下面是这个例子的具体运算方式：

object SourceQueueDemo extends App {
  implicit val sys = ActorSystem("demoSys")
  implicit val ec = sys.dispatcher
  implicit val mat = ActorMaterializer(
    ActorMaterializerSettings(sys)
      .withInputBuffer(initialSize = 16, maxSize = 16)
  )
  
  val source: Source[String, SourceQueueWithComplete[String]]  =
               Source.queue[String](bufferSize = 16,
               overflowStrategy = OverflowStrategy.backpressure)

  val inputQueue: SourceQueueWithComplete[String] = source.toMat(Sink.foreach(println))(Keep.left).run()

  inputQueue.watchCompletion().onComplete {
    case Success(result) => println(s"Calculator ends with: $result")
    case Failure(cause)  => println(s"Calculator ends with exception: ${cause.getMessage}")
  }

  val calc = sys.actorOf(Calculator.props(inputQueue),"calculator")

  import Calculator._
  
  calc ! Add(3,5)
  scala.io.StdIn.readLine
  calc ! Add(39,1)
  scala.io.StdIn.readLine
  calc ! ProduceError(new Exception("Boooooommm!"))
  scala.io.StdIn.readLine
  calc ! Add(1,1)

  scala.io.StdIn.readLine
  sys.terminate()

}

在本次讨论里我们了解了akka-stream与外界系统对接集成的一些情况。主要介绍了一些支持Reactive-Stream backpressure的方法。

以下是本次示范的全部源代码：

MapAsyncDemo.scala:

import akka.actor._
import akka.pattern._
import akka.stream._
import akka.stream.scaladsl._
import akka.routing._

import scala.concurrent.duration._
import akka.util.Timeout

object StorageActor {

  case class Query(rec: Int, qry: String) //模拟数据存写Query
  class DbException(cause: String) extends Exception(cause) //自定义存写异常

  class StorageActor extends Actor with ActorLogging { //存写操作Actor
    override def receive: Receive = {
      case Query(num,qry) =>
        var res: String = ""
        try {
          res = saveToDB(num,qry)
        } catch {
          case e: Exception => Error(num,qry) //模拟操作异常
        }
        sender() ! res
      case _ =>
    }
    def saveToDB(num: Int,qry: String): String = { //模拟存写函数
      val msg = s"${self.path} is saving: [$qry#$num]"
      if ( num % 5 == 0) Error(num,qry)        //模拟异常
      else {
        log.info(s"${self.path} is saving: [$qry#$num]")
        s"${self.path} is saving: [$qry#$num]"
      }
    }
    def Error(num: Int,qry: String): String = {
      val msg = s"${self.path} is saving: [$qry#$num]"
      sender() ! msg                       //卸去backpressure
      throw new DbException(s"$msg blew up, boooooom!!!")
    }

    //验证异常重启
    //BackoffStrategy.onStop goes through restart process
    override def preRestart(reason: Throwable, message: Option[Any]): Unit = {
      log.info(s"Restarting ${self.path.name} on ${reason.getMessage}")
       message match {
        case Some(Query(n,qry)) =>
          self ! Query(n+101,qry)      //把异常消息再补发送给自己，n+101更正了异常因素
        case _ =>
          log.info(s"Exception message: None")

      }
      super.preRestart(reason, message)
    }

    override def postRestart(reason: Throwable): Unit = {
      log.info(s"Restarted ${self.path.name} on ${reason.getMessage}")
      super.postRestart(reason)
    }

    override def postStop(): Unit = {
      log.info(s"Stopped ${self.path.name}!")
      super.postStop()
    }
   //BackOffStrategy.onFailure dosn‘t go through restart process
    override def preStart(): Unit = {
      log.info(s"PreStarting ${self.path.name} ...")
      super.preStart()
    }
  }
  def props = Props(new StorageActor)
}

object StorageActorGuardian {  //带监管策略的StorageActor
  def props: Props = { //在这里定义了监管策略和StorageActor构建
    def decider: PartialFunction[Throwable, SupervisorStrategy.Directive] = {
      case _: StorageActor.DbException => SupervisorStrategy.Restart
    }

    val options = Backoff.onStop(StorageActor.props, "dbWriter", 100 millis, 500 millis, 0.0)
      .withManualReset
      .withSupervisorStrategy(
        OneForOneStrategy(maxNrOfRetries = 3, withinTimeRange = 1 second)(
          decider.orElse(SupervisorStrategy.defaultDecider)
        )
      )
    BackoffSupervisor.props(options)
  }
}

object IntegrateDemo extends App {
  implicit val sys = ActorSystem("demoSys")
  implicit val ec = sys.dispatcher
  implicit val mat = ActorMaterializer(
    ActorMaterializerSettings(sys)
      .withInputBuffer(initialSize = 16, maxSize = 16)
  )
  
  val numOfActors = 3
  val routees: List[ActorRef] = List.fill(numOfActors)(
    sys.actorOf(StorageActorGuardian.props))
  val routeePaths: List[String] = routees.map{ref => "/user/"+ref.path.name}

  val storageActorPool = sys.actorOf(
    RoundRobinGroup(routeePaths).props()
      .withDispatcher("akka.pool-dispatcher")
    ,"starageActorPool"
  )
  
  implicit val timeout = Timeout(3 seconds)
  Source(Stream.from(1)).delay(3.second,DelayOverflowStrategy.backpressure)
    .mapAsync(parallelism = 1){ n =>
      (storageActorPool ? StorageActor.Query(n,s"Record")).mapTo[String]
    }.runWith(Sink.foreach(println))

  scala.io.StdIn.readLine()
  sys.terminate()

}

SinkActorRefAckDemo.scala:

package sinkactorrefack
import akka.actor._
import akka.pattern._
import akka.stream._
import akka.stream.scaladsl._
import akka.routing._

import scala.concurrent.duration._

object StorageActor {
  val onInitMessage = "start"
  val onCompleteMessage = "done"
  val ackMessage = "ack"

  case class Query(rec: Int, qry: String) //模拟数据存写Query
  class DbException(cause: String) extends Exception(cause) //自定义存写异常

  class StorageActor extends Actor with ActorLogging { //存写操作Actor
    override def receive: Receive = {
      case `onInitMessage` => sender() ! ackMessage
      case Query(num,qry) =>
        var res: String = ""
        try {
          res = saveToDB(num,qry)
        } catch {
          case e: Exception => Error(num,qry) //模拟操作异常
        }
        sender() ! ackMessage
      case `onCompleteMessage` => //clean up resources 释放资源
      case _ =>
    }
    def saveToDB(num: Int,qry: String): String = { //模拟存写函数
      val msg = s"${self.path} is saving: [$qry#$num]"
      if ( num == 3) Error(num,qry)        //模拟异常
      else {
        log.info(s"${self.path} is saving: [$qry#$num]")
        s"${self.path} is saving: [$qry#$num]"
      }
    }
    def Error(num: Int,qry: String) = {
      val msg = s"${self.path} is saving: [$qry#$num]"
      sender() ! ackMessage
      throw new DbException(s"$msg blew up, boooooom!!!")
    }

    //验证异常重启
    //BackoffStrategy.onStop goes through restart process
    override def preRestart(reason: Throwable, message: Option[Any]): Unit = {
      log.info(s"Restarting ${self.path.name} on ${reason.getMessage}")
      message match {
        case Some(Query(n,qry)) =>
          self ! Query(n+101,qry)      //把异常消息再补发送给自己，n+101更正了异常因素
        case _ =>
          log.info(s"Exception message: None")

      }
      super.preRestart(reason, message)
    }

    override def postRestart(reason: Throwable): Unit = {
      log.info(s"Restarted ${self.path.name} on ${reason.getMessage}")
      super.postRestart(reason)
    }

    override def postStop(): Unit = {
      log.info(s"Stopped ${self.path.name}!")
      super.postStop()
    }
    //BackOffStrategy.onFailure dosn‘t go through restart process
    override def preStart(): Unit = {
      log.info(s"PreStarting ${self.path.name} ...")
      super.preStart()
    }
  }
  def props = Props(new StorageActor)
}

object StorageActorGuardian {  //带监管策略的StorageActor
  def props: Props = { //在这里定义了监管策略和StorageActor构建
    def decider: PartialFunction[Throwable, SupervisorStrategy.Directive] = {
      case _: StorageActor.DbException => SupervisorStrategy.Restart
    }

    val options = Backoff.onStop(StorageActor.props, "dbWriter", 100 millis, 500 millis, 0.0)
      .withManualReset
      .withSupervisorStrategy(
        OneForOneStrategy(maxNrOfRetries = 3, withinTimeRange = 1 second)(
          decider.orElse(SupervisorStrategy.defaultDecider)
        )
      )
    BackoffSupervisor.props(options)
  }
}

object SinkActorRefWithAck extends App {
  import StorageActor._

  implicit val sys = ActorSystem("demoSys")
  implicit val ec = sys.dispatcher
  implicit val mat = ActorMaterializer(
    ActorMaterializerSettings(sys)
      .withInputBuffer(initialSize = 16, maxSize = 16)
  )

  val storageActor = sys.actorOf(StorageActor.props,"storageActor")

  val numOfActors = 3
  val routees: List[ActorRef] = List.fill(numOfActors)(
    sys.actorOf(StorageActorGuardian.props))
  val routeePaths: List[String] = routees.map{ref => "/user/"+ref.path.name}

  val storageActorPool = sys.actorOf(
    RoundRobinGroup(routeePaths).props()
      .withDispatcher("akka.pool-dispatcher")
    ,"starageActorPool"
  )

  Source(Stream.from(1)).map(n => Query(n,s"Record")).delay(3.second,DelayOverflowStrategy.backpressure)
      .runWith(Sink.actorRefWithAck(
       storageActorPool, onInitMessage, ackMessage,onCompleteMessage))

  scala.io.StdIn.readLine()
  sys.terminate()

}

SourceQueueDemo.scala：

import akka.actor._
import akka.stream._
import akka.stream.scaladsl._
import scala.concurrent._
import scala.util._
import akka.pattern._

object Calculator {
  trait Operations
  case class Add(op1:Int, op2:Int) extends Operations
  case class DisplayError(err: Exception) extends Operations
  case object Stop extends Operations
  case class ProduceError(err: Exception) extends Operations

  def props(inputQueue: SourceQueueWithComplete[String]) = Props(new Calculator(inputQueue))
}
class Calculator(inputQueue: SourceQueueWithComplete[String]) extends Actor with ActorLogging{
  import Calculator._
  import context.dispatcher
  override def receive: Receive = {
    case Add(op1,op2) =>
      val msg = s"$op1 + $op2 = ${op1 + op2}"
      inputQueue.offer(msg)
        .recover {
        case e: Exception => DisplayError(e)}
        .pipeTo(self).mapTo[String]
    case QueueOfferResult =>
      log.info("QueueOfferResult.Enqueued")
    case QueueOfferResult.Enqueued =>
      log.info("QueueOfferResult.Enqueued")
    case QueueOfferResult.Dropped =>
    case QueueOfferResult.Failure(cause) =>
    case QueueOfferResult.QueueClosed  =>
      log.info("QueueOfferResult.QueueClosed")

    case Stop => inputQueue.complete()
    case ProduceError(e) => inputQueue.fail(e)

  }
}


object SourceQueueDemo extends App {
  implicit val sys = ActorSystem("demoSys")
  implicit val ec = sys.dispatcher
  implicit val mat = ActorMaterializer(
    ActorMaterializerSettings(sys)
      .withInputBuffer(initialSize = 16, maxSize = 16)
  )

  val source: Source[String, SourceQueueWithComplete[String]]  =
               Source.queue[String](bufferSize = 16,
               overflowStrategy = OverflowStrategy.backpressure)

  val inputQueue: SourceQueueWithComplete[String] = source.toMat(Sink.foreach(println))(Keep.left).run()

  inputQueue.watchCompletion().onComplete {
    case Success(result) => println(s"Calculator ends with: $result")
    case Failure(cause)  => println(s"Calculator ends with exception: ${cause.getMessage}")
  }

  val calc = sys.actorOf(Calculator.props(inputQueue),"calculator")

  import Calculator._

  calc ! Add(3,5)
  scala.io.StdIn.readLine
  calc ! Add(39,1)
  scala.io.StdIn.readLine
  calc ! ProduceError(new Exception("Boooooommm!"))
  scala.io.StdIn.readLine
  calc ! Add(1,1)

  scala.io.StdIn.readLine
  sys.terminate()

}