org.apache.spark.SparkException:作业由于阶段故障而中止 - OOM 异常

Question

在我的应用程序中，我使用如下所示的 Spark 分区挖出一个包含 500 万行和 151 列的表，并将其持久化为DISK_ONLY

   val query = "(select * from destinationlarge) as dest"
val options = Map(
  "url" -> "jdbc:mysql://IPADDRESS:3306/test?useSSL=false",
  "driver" -> "com.mysql.jdbc.Driver",
  "dbtable" -> query,
  "user" -> "root",
  "password" -> "root")
val destination = spark.read.options(options).jdbc(options("url"), options("dbtable"), "0", 1, 5, 4, new java.util.Properties()).rdd.map(_.mkString(",")).persist(StorageLevel.DISK_ONLY)

集群有 5 个数据节点和 1 个硬件配置 i3 的名称节点，每个 4 个内核和 4 GB RAM，在执行一段时间后，其中一个执行程序已死并抛出以下错误

Exception in thread "main" org.apache.spark.SparkException: Job aborted due to stage failure: Task 0 in stage 0.0 failed 4 times, most recent failure: Lost task 0.3 in stage 0.0 (TID 6, datanode5, executor 6): ExecutorLostFailure (executor 6 exited caused by one of the running tasks) Reason: Executor heartbeat timed out after 139401 ms
Driver stacktrace:
at org.apache.spark.scheduler.DAGScheduler.org$apache$spark$scheduler$DAGScheduler$$failJobAndIndependentStages(DAGScheduler.scala:1435)
at org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1423)
at org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1422)
at scala.collection.mutable.ResizableArray$class.foreach(ResizableArray.scala:59)
at scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:48)
at org.apache.spark.scheduler.DAGScheduler.abortStage(DAGScheduler.scala:1422)
at org.apache.spark.scheduler.DAGScheduler$$anonfun$handleTaskSetFailed$1.apply(DAGScheduler.scala:802)
at org.apache.spark.scheduler.DAGScheduler$$anonfun$handleTaskSetFailed$1.apply(DAGScheduler.scala:802)
at scala.Option.foreach(Option.scala:257)
at org.apache.spark.scheduler.DAGScheduler.handleTaskSetFailed(DAGScheduler.scala:802)
at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.doOnReceive(DAGScheduler.scala:1650)
at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:1605)
at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:1594)
at org.apache.spark.util.EventLoop$$anon$1.run(EventLoop.scala:48)
at org.apache.spark.scheduler.DAGScheduler.runJob(DAGScheduler.scala:628)
at org.apache.spark.SparkContext.runJob(SparkContext.scala:1925)
at org.apache.spark.SparkContext.runJob(SparkContext.scala:1938)
at org.apache.spark.SparkContext.runJob(SparkContext.scala:1951)
at org.apache.spark.SparkContext.runJob(SparkContext.scala:1965)
at org.apache.spark.rdd.RDD.count(RDD.scala:1158)
at com.syntel.spark.sparkDVT$.main(sparkDVT.scala:68)
at com.syntel.spark.sparkDVT.main(sparkDVT.scala)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:497)
at org.apache.spark.deploy.SparkSubmit$.org$apache$spark$deploy$SparkSubmit$$runMain(SparkSubmit.scala:750)
at org.apache.spark.deploy.SparkSubmit$.doRunMain$1(SparkSubmit.scala:187)
at org.apache.spark.deploy.SparkSubmit$.submit(SparkSubmit.scala:212)
at org.apache.spark.deploy.SparkSubmit$.main(SparkSubmit.scala:126)
at org.apache.spark.deploy.SparkSubmit.main(SparkSubmit.scala)

lowerbound=1 、upperbound=5 、number of partitions is 4此链接 （https://www.dezyre.com/article/how-data-partitioning-in-spark-helps-achieve-more-parallelism/297） 中建议的内核总数等于分区数，即所有节点中的 4 个内核，因此 4 个分区。

spark-submit

spark-submit --class "com.syntel.spark.sparkDVT" --master yarn --jars --executor-memory 512m --executor-cores 1 --num-executors 5 /root/sparkdvtmysql_2.11-1.0.jar

如果我错了，请纠正我

谢谢

Answer 1

我建议您按原样使用DataFame（在Spark 2.0中，即DataSet[Row]），因为DataSet使用编码器，因此它的内存占用比RDD少。

val destination = spark.read
    .options(options)
    .format("jdbc")
    .load()

如果你想通过分隔符连接列，你可以使用 concat_ws（） - 这里的例子

destination
  .withColumn("column", concat_ws(", ", 
     destination.columns.map(destination.col(_)).toSeq : _*))
  .select("id, column") // id will be used for subtraction with other df
  .persist(StorageLevel.DISK_ONLY)

查看这篇文章 - 比较RDD/DF/DS，您了解数据集与RDD的不同之处及其优势。

这可能无法完全回答您的问题。我会根据我的评论回复更新 asnwer