[SPARK-32813] Reading parquet rdd in non columnar mode fails in multithreaded environment

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.0.0, 3.1.0
• Fix Version/s: 3.0.2, 3.1.0
• Component/s: SQL

Environment

Spark 3.0.0, Scala 2.12.12

• Assignee: L. C. Hsieh
• Reporter: Vladimir Klyushnikov
• Created: 07/Sep/20 16:31
• Updated: 12/Dec/22 18:10
• Resolved: 09/Sep/20 03:24

Description

Reading parquet rdd in non columnar mode (i.e. with list fields) if Spark session was  created in one thread and rdd is being read in another  - so InheritableThreadLocal  with active session is not propagated. Code below was working perfectly in Spark 2.X, but fails in Spark 3

import java.util.concurrent.Executors

import org.apache.spark.sql.SparkSession

import scala.concurrent.{Await, ExecutionContext, Future}
import scala.concurrent.duration._

object Main {

  final case class Data(list: List[Int])

  def main(args: Array[String]): Unit = {

    val executor1 = Executors.newSingleThreadExecutor()
    val executor2 = Executors.newSingleThreadExecutor()
    try {
      val ds = Await.result(Future {
        val session = SparkSession.builder().appName("test").master("local[*]").getOrCreate()
        import session.implicits._

        val path = "test.parquet"
        session.createDataset(Data(1 :: Nil) :: Nil).write.parquet(path)
        session.read.parquet(path).as[Data]
      }(ExecutionContext.fromExecutorService(executor1)), 1.minute)

      Await.result(Future {
        ds.rdd.collect().foreach(println(_))
      }(ExecutionContext.fromExecutorService(executor2)), 1.minute)

    } finally {
      executor1.shutdown()
      executor2.shutdown()
    }
  }
}

This code fails with following exception:

Exception in thread "main" java.util.NoSuchElementException: None.getException in thread "main" java.util.NoSuchElementException: None.get at scala.None$.get(Option.scala:529) at scala.None$.get(Option.scala:527) at org.apache.spark.sql.execution.FileSourceScanExec.needsUnsafeRowConversion$lzycompute(DataSourceScanExec.scala:178) at org.apache.spark.sql.execution.FileSourceScanExec.needsUnsafeRowConversion(DataSourceScanExec.scala:176) at org.apache.spark.sql.execution.FileSourceScanExec.doExecute(DataSourceScanExec.scala:462) at org.apache.spark.sql.execution.SparkPlan.$anonfun$execute$1(SparkPlan.scala:175) at org.apache.spark.sql.execution.SparkPlan.$anonfun$executeQuery$1(SparkPlan.scala:213) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151) at org.apache.spark.sql.execution.SparkPlan.executeQuery(SparkPlan.scala:210) at org.apache.spark.sql.execution.SparkPlan.execute(SparkPlan.scala:171) at org.apache.spark.sql.execution.DeserializeToObjectExec.doExecute(objects.scala:96) at org.apache.spark.sql.execution.SparkPlan.$anonfun$execute$1(SparkPlan.scala:175) at org.apache.spark.sql.execution.SparkPlan.$anonfun$executeQuery$1(SparkPlan.scala:213) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151) at org.apache.spark.sql.execution.SparkPlan.executeQuery(SparkPlan.scala:210) at org.apache.spark.sql.execution.SparkPlan.execute(SparkPlan.scala:171) at org.apache.spark.sql.execution.QueryExecution.toRdd$lzycompute(QueryExecution.scala:122) at org.apache.spark.sql.execution.QueryExecution.toRdd(QueryExecution.scala:121) at org.apache.spark.sql.Dataset.rdd$lzycompute(Dataset.scala:3198) at org.apache.spark.sql.Dataset.rdd(Dataset.scala:3196)

Issue Links

• is duplicated by: SPARK-32589 NoSuchElementException: None.get for needsUnsafeRowConversion
• links to: [Github] Pull Request #29667 (viirya)