Akka Singleton-不接受邮件

是我太愚蠢了——我没有把索引器道具传递到系统创建中。我会把答案留在这里，以防有人从中受益*

我正在创建一个singleton并发送如下消息：

 val indexerProps = ClusterSingletonManager.props(had => Props(
            classOf[SingleCoreIndexer], dataProvider, publisher, name), name, End, None)

        val coreIndexer = system.actorOf(indexerProps, name)
        //val coreIndexer = system.actorOf(Props(classOf[SingleCoreIndexer], dataProvider, publisher, name))

        coreIndexer ! "start_indexing"

注释掉的行显示了工作正常的非单例道具

当我运行应用程序时，会出现以下错误：

[WARN] [06/21/2013 11:55:32.443] [deadcoreindexerstest-akka.actor.default-dispatcher-5] [akka://deadcoreindexerstest/user/node1] unhandled event start_indexing in state Start

所有其他功能都停止工作，这与暗示“coreIndexer”参与者没有获得“start\u indexing”消息的消息相关

更多代码：

class Indexer(systemCreator: SystemCreator, publisherProps: Props, dataProviderProps: Props, name: String) {

    def start {
        val system = systemCreator.create
        val dataProvider = system.actorOf(dataProviderProps)
        val publisher = system.actorOf(publisherProps)

        val indexerProps = ClusterSingletonManager.props(
            singletonProps = had => Props(classOf[SingleCoreIndexer], dataProvider, publisher, name),
            singletonName = "aaa",
            terminationMessage = End,
            role = None
            )

        val coreIndexer = system.actorOf(Props(classOf[SingleCoreIndexer], dataProvider, publisher, name))
        coreIndexer ! "start_indexing"
    }
}



class SingleCoreIndexer(dataProvider: ActorRef, publisher: ActorRef, name: String) extends Actor {

    def receive = {

        case "start_indexing" => {
            println("Single core indexer starting indexing")
            dataProvider ! new NextBatchOfDataPlease
        }

        case BatchOfData(data) => {
            publisher ! (name, data)
            self ! "next_batch"
        }

        case "next_batch" => {
            dataProvider ! new NextBatchOfDataPlease
        }
    }
}

看起来我是在给经理发信息，而不是给单身汉。但是，当我向singleton发送消息时，什么也没有发生：

class Indexer(systemCreator: SystemCreator, publisherProps: Props, dataProviderProps: Props, name: String) {

    def start {
        val system = systemCreator.create
        val dataProvider = system.actorOf(dataProviderProps)
        val publisher = system.actorOf(publisherProps)

        val indexerProps = ClusterSingletonManager.props(
            singletonProps = had => Props(classOf[SingleCoreIndexer], dataProvider, publisher, name),
            singletonName = "singlecoreindexer",
            terminationMessage = End,
            role = None
            )

        system.actorOf(Props(classOf[SingleCoreIndexer], dataProvider, publisher, name))
        val coreIndexer = system.actorSelection(s"/user/$name/singlecoreindexer")
        coreIndexer ! "start_indexing"
    }
}

---

你看到的问题（我认为）源于你向

ClusterSingletonManager

发送了一条消息，而不是发送给坐在它下面的实际参与者。试着按姓名（

actor或

）查找下面的参与者，它应该会起作用。

我知道这个问题已经解决了，但即使有了这里提供的信息，我还是花了一段时间才弄清楚如何向单身汉发送消息，并决定留下我在这里找到的信息

文档中没有很好解释的两个关键概念是：

使用

ClusterSingletonManager.props创建的参与者是实际实例的父级，并且

您应该仅使用该参与者的地址来创建

ClusterSingletonProxy

集群中的每个节点都将创建一个单例管理器，最终胜出的最老节点是您实际想要与之交谈的单例的父节点

ClusterSingletonProxy

确保您正在与实际的代理对话，并且即使singleton暂时不可用或迁移到另一个节点，您也始终在与适当的实例对话

鉴于这些信息，代码应为：

class Indexer(systemCreator: SystemCreator, publisherProps: Props, dataProviderProps: Props, name: String) {

  def start {
    val system = systemCreator.create
    val dataProvider = system.actorOf(dataProviderProps)
    val publisher = system.actorOf(publisherProps)

    val indexerProps = ClusterSingletonManager.props(
      singletonProps = had => Props(classOf[SingleCoreIndexer], dataProvider, publisher, name),
      singletonName = "singlecoreindexer",
      terminationMessage = End,
      role = None
    )

    val singletonManager = system.actorOf(
      Props(classOf[SingleCoreIndexer],dataProvider, publisher, name)
    )

    val indexerPath = (singletonManager.path / name)
    val coreIndexer = system.actorOf(
      ClusterSingletonProxy.props(indexerPath, None),
      s"$name-proxy"
    )

    coreIndexer ! "start_indexing"
  }
}

---

您能否共享

SingleCoreIndexer

的代码？看起来这是某种与FSM相关的问题。有趣的是，在我提交我的答案时，你得到了你的答案，这是相同的结论。使用

ClusterSingletonProxy

访问singleton manager后面的actor。