[jira] [Created] (SPARK-38137) Repartition+Shuffle+ non deterministic function leads to bad results

["Jakub Leś (Jira)" <ji...@apache.org>]

Jakub Leś created SPARK-38137:
---------------------------------

             Summary: Repartition+Shuffle+ non deterministic function leads to bad results
                 Key: SPARK-38137
                 URL: https://issues.apache.org/jira/browse/SPARK-38137
             Project: Spark
          Issue Type: Bug
          Components: Spark Core
    Affects Versions: 3.2.1, 3.1.1
            Reporter: Jakub Leś


Hi,

The results when using a non deterministic function in repartition (like rand) leads into incorrect results.

Reproduce: (correct)

 
{code:java}
// code placeholder
import scala.sys.process._
import org.apache.spark.TaskContext
import org.apache.spark.sql.functions.randval res = spark.range(0, 100 * 100, 1).repartition(200).map { x =>
  x
}.repartition(200).map { x =>
  if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 2) {
    throw new Exception("pkill -f java".!!)
  }
  x
}
res.distinct().count() {code}
The correct result 10000 

Reproduce: (bad)

 
{code:java}
// code placeholder
import scala.sys.process._
import org.apache.spark.TaskContext
import org.apache.spark.sql.functions.randval res = spark.range(0, 100 * 100, 1).repartition(200).map { x =>
  x
}.repartition(10, Array(rand):_*).map { x =>
  if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 2) {
    throw new Exception("pkill -f java".!!)
  }
  x
}
res.distinct().count() {code}
The bad result 9396 

 



--
This message was sent by Atlassian Jira
(v8.20.1#820001)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org