简体   繁体   English

在 Spark 中为每个执行程序创建单例对象

[英]Creating singleton object in Spark for each executor

I've quite a similar problem described here: How to perform one operation on each executor once in spark I followed the first approach in the first answer but still got a Serialization problem.我在这里描述了一个非常类似的问题: How to perform an operation on each executor once in spark我遵循了第一个答案中的第一种方法,但仍然遇到了序列化问题。

What I want to do is, I have queries like a tuple of (sourceVertex, targetVertex) and send these to executors and executors will return to me a shortest path.我想要做的是,我有像 (sourceVertex, targetVertex) 元组这样的查询,并将它们发送给执行程序,执行程序将返回给我一条最短路径。 For this I'm using jgrapht.为此,我正在使用 jgrapht。

When I implement like this当我像这样实施时

class ShortestPath(graph: SimpleDirectedWeightedGraph[Node, DefaultWeightedEdge], 
bc: Broadcast[SimpleDirectedWeightedGraph[Node, DefaultWeightedEdge]]) extends Serializable {

  def calculateShortestPath(vertexRDD: RDD[Node]) = {

    val result = vertexRDD.map(vertex => {
      val dijkstraShortestPath: DijkstraShortestPath[Node, DefaultWeightedEdge] 
                = new DijkstraShortestPath[Node, DefaultWeightedEdge](bc.value)
      val distanceIn = dijkstraShortestPath.getPath(vertex, Node(4, 1, true)).getWeight()
      distanceIn
    })
    result.collect().foreach(println(_))
  }

}

object ShortestPath {
  def apply(graph: SimpleDirectedWeightedGraph[Node, DefaultWeightedEdge], 
bc: Broadcast[SimpleDirectedWeightedGraph[Node, DefaultWeightedEdge]]): ShortestPath = new ShortestPath(graph, bc)
}

Everything is okay But the problem is I think I'm creating dijkstraShortestPath object for each task, am I right?一切正常但问题是我想我正在为每个任务创建dijkstraShortestPath对象,对吗?

My aim is to create this object for each executor and use it for each task on that executor.我的目标是为每个执行程序创建此对象,并将其用于该执行程序上的每个任务。

The link that I gave says create a Object with lazy val, instantiate your think here then use it RDD map function.我给出的链接说用惰性值创建一个对象,在这里实例化你的想法,然后使用它 RDD 映射函数。 I implement that solution like this:我实现了这样的解决方案:

object Dij {
  lazy val dijsktra = {
    val graph = GraphCreator.createGraph()
    val dijkstraShortestPath: DijkstraShortestPath[Node, DefaultWeightedEdge] = new DijkstraShortestPath[Node, DefaultWeightedEdge](graph)
    dijkstraShortestPath
  }
}

and used in ShortestPath class并在 ShortestPath 类中使用

    val result = vertexRDD.map(vertex => {
      val dijkstraShortestPath = Dij.dijsktra
      val distanceIn = dijkstraShortestPath.getPath(vertex, Node(4, 1, true)).getWeight()
      dijkstraShortestPath
    })

    result.collect().foreach(println(_))

but then I'm getting serialization error thank says - object not serializable (class: org.jgrapht.alg.shortestpath.DijkstraShortestPath, value: org.jgrapht.alg.shortestpath.DijkstraShortestPath@2cb8e13b) which is right, when i look implementation there is no Serializable.但后来我收到序列化错误谢谢说- object not serializable (class: org.jgrapht.alg.shortestpath.DijkstraShortestPath, value: org.jgrapht.alg.shortestpath.DijkstraShortestPath@2cb8e13b)这是正确的,当我在那里查看实现时是不可序列化的。

And another question is if it is not Serializable then how my first implementation worked?另一个问题是如果它不是 Serializable 那么我的第一个实现是如何工作的?

I think there is an unintended error in your second snippet.我认为您的第二个代码段中存在意外错误。

The first function given to a map returns a weight (presumably a Double ?) the second returns a DijkstraShortestPath which is not serializable.赋予地图的第一个函数返回一个权重(大概是一个Double ?),第二个函数返回一个不可序列化的DijkstraShortestPath This explains why the two snippets behave differently.这解释了为什么这两个片段的行为不同。

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM