[SPARK-7103] SparkContext.union crashed when some RDDs have no partitioner

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.3.0, 1.3.1
• Fix Version/s: 1.3.2, 1.4.0
• Component/s: Spark Core
• Target Version/s: 1.3.2, 1.4.0
• Assignee: Steven She
• Reporter: Steven She
• Created: 23/Apr/15 21:24
• Updated: 27/Apr/15 22:56
• Resolved: 27/Apr/15 22:55

Description

I encountered a bug where Spark crashes with the following stack trace:

java.util.NoSuchElementException: None.get
	at scala.None$.get(Option.scala:313)
	at scala.None$.get(Option.scala:311)
	at org.apache.spark.rdd.PartitionerAwareUnionRDD.getPartitions(PartitionerAwareUnionRDD.scala:69)

Here's a minimal example that reproduces it on the Spark shell:

val x = sc.parallelize(Seq(1->true,2->true,3->false)).partitionBy(new HashPartitioner(1))
val y = sc.parallelize(Seq(1->true))
sc.union(y, x).count() // crashes
sc.union(x, y).count() // This works since the first RDD has a partitioner

We had to resort to instantiating the UnionRDD directly to avoid the PartitionerAwareUnionRDD.

Issue Links

• links to: [Github] Pull Request #5678 (vinodkc)
• links to: [Github] Pull Request #5679 (stevencanopy)