从Spark驱动程序反序列化Java的本机readObject时

Question

我有两个火花作业A和B，因此A必须在B之前运行。A的输出必须可从以下位置读取：

• 火花工作B
• Spark环境之外的独立Scala程序（没有Spark依赖）

我目前正在使用Java的本机序列化与Scala案例类。

从“火花工作”开始：

val model = ALSFactorizerModel(...)

context.writeSerializable(resultOutputPath, model)

使用序列化方法：

def writeSerializable[T <: Serializable](path: String, obj: T): Unit = {
  val writer: OutputStream = ... // Google Cloud Storage dependant
  val oos: ObjectOutputStream = new ObjectOutputStream(writer)
  oos.writeObject(obj)
  oos.close()
  writer.close()
}

从B Spark作业或任何独立的非Spark Scala代码：

val lastFactorizerModel: ALSFactorizerModel = context
                     .readSerializable[ALSFactorizerModel](ALSFactorizer.resultOutputPath)

使用反序列化方法：

def readSerializable[T <: Serializable](path: String): T = {
  val is : InputStream = ... // Google Cloud Storage dependant
  val ois = new ObjectInputStream(is)
  val model: T = ois
    .readObject()
    .asInstanceOf[T]
  ois.close()
  is.close()

  model
}

（嵌套的）案例类：

模型：

package mycompany.algo.als.common.io.model.factorizer

import mycompany.data.item.ItemStore

@SerialVersionUID(1L)
final case class ALSFactorizerModel(
  knownItems:       Array[ALSFeaturedKnownItem],
  unknownItems:     Array[ALSFeaturedUnknownItem],
  rank:             Int,
  modelTS:          Long,
  itemRepositoryTS: Long,
  stores:           Seq[ItemStore]
) {   
}

项目存储：

package mycompany.data.item

@SerialVersionUID(1L)
final case class ItemStore(
  id:     String,
  tenant: String,
  name:   String,
  index:  Int
) {
}

输出：

• 来自独立的非Spark Scala程序=

例外情况：

java.lang.ClassCastException: cannot assign instance of scala.collection.immutable.List$SerializationProxy to field mycompany.algo.als.common.io.model.factorizer.ALSFactorizerModel.stores of type scala.collection.Seq in instance of mycompany.algo.als.common.io.model.factorizer.ALSFactorizerModel
  at java.io.ObjectStreamClass$FieldReflector.setObjFieldValues(ObjectStreamClass.java:2133)
  at java.io.ObjectStreamClass.setObjFieldValues(ObjectStreamClass.java:1305)
  at java.io.ObjectInputStream.defaultReadFields(ObjectInputStream.java:2251)
  at java.io.ObjectInputStream.readSerialData(ObjectInputStream.java:2169)
  at java.io.ObjectInputStream.readOrdinaryObject(ObjectInputStream.java:2027)
  at java.io.ObjectInputStream.readObject0(ObjectInputStream.java:1535)
  at java.io.ObjectInputStream.readObject(ObjectInputStream.java:422)
  at mycompany.fs.gcs.SimpleGCSFileSystem.readSerializable(SimpleGCSFileSystem.scala:71)
  at mycompany.algo.als.batch.strategy.ALSClusterer$.run(ALSClusterer.scala:38)
  at mycompany.batch.SinglePredictorEbapBatch$$anonfun$3.apply(SinglePredictorEbapBatch.scala:55)
  at mycompany.batch.SinglePredictorEbapBatch$$anonfun$3.apply(SinglePredictorEbapBatch.scala:55)
  at scala.concurrent.impl.Future$PromiseCompletingRunnable.liftedTree1$1(Future.scala:24)
  at scala.concurrent.impl.Future$PromiseCompletingRunnable.run(Future.scala:24)
  at scala.concurrent.impl.ExecutionContextImpl$AdaptedForkJoinTask.exec(ExecutionContextImpl.scala:121)
  at scala.concurrent.forkjoin.ForkJoinTask.doExec(ForkJoinTask.java:260)
  at scala.concurrent.forkjoin.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:1339)
  at scala.concurrent.forkjoin.ForkJoinPool.runWorker(ForkJoinPool.java:1979)
  at scala.concurrent.forkjoin.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:107)

我是不是漏了什么？我应该配置Dataproc/Spark以支持对该代码使用Java序列化吗？

我用--jars提交作业

Scala版本：2.11.8 Spark版本：2.0.2 SBT版本：0.13.13

谢谢你的帮助




Answer 1

将stores:Seq[ItemStore]替换为stores:Array[ItemStore]为我们解决了这个问题。

或者，我们可以使用另一个类加载器来执行ser/deser-iize操作。

希望这能有所帮助。