Spark：对象不可序列化

Question

我有一个批处理作业，我尝试将其转换为结构化流。我收到以下错误：

20/03/31 15:09:23 ERROR Executor: Exception in task 0.0 in stage 0.0 (TID 0)
java.io.NotSerializableException: com.apple.ireporter.analytics.compute.AggregateKey
Serialization stack:
    - object not serializable (class: com.apple.ireporter.analytics.compute.AggregateKey, value: d_)

...其中“ d_”是数据集中的最后一行

这是相关的代码片段

    df.writeStream.foreachBatch { (batchDF: DataFrame, batchId: Long) =>
      import spark.implicits._
      val javaRdd = batchDF.toJavaRDD
      val dataframeToRowColFunction = new RowToColumn(table)
      println("Back to Main class")
      val combinedRdd =javaRdd.flatMapToPair(dataframeToRowColFunction.FlatMapData2).combineByKey(aggrCreateComb.createCombiner,aggrMerge.aggrMerge,aggrMergeCombiner.aggrMergeCombiner)
      // spark.createDataFrame( combinedRdd).show(1); // I commented this
      // combinedRdd.collect() // I added this as a test
    }

这是FlatMapData2类

  val FlatMapData2: PairFlatMapFunction[Row, AggregateKey, AggregateValue] = new PairFlatMapFunction[Row, AggregateKey, AggregateValue]() {
    //val FlatMapData: PairFlatMapFunction[Row, String, AggregateValue] = new PairFlatMapFunction[Row, String, AggregateValue]() {
    override def call(x: Row) = {
      val tuples = new util.ArrayList[Tuple2[AggregateKey, AggregateValue]]
      val decomposedEvents = decomposer.decomposeDistributed(x)
      decomposedEvents.foreach {
        y =>  tuples.add(Tuple2(y._1,y._2))
      }
      tuples.iterator()
    }
  }

这里是聚合键类

class AggregateKey(var partitionkeys: Map[Int,Any],var clusteringkeys : Map[Int,Any]) extends Comparable [AggregateKey]{
...
}

我对此并不陌生，我们将不胜感激。请让我知道是否需要添加其他内容

Answer 1