Andy Grove created SPARK-34682:
----------------------------------
Summary: Regression in "operating on canonicalized plan" check in
CustomShuffleReaderExec
Key: SPARK-34682
URL: https://issues.apache.org/jira/browse/SPARK-34682
Project: Spark
Issue Type: Bug
Components: SQL
Affects Versions: 3.1.1
Reporter: Andy Grove
Fix For: 3.2.0, 3.1.2
In Spark 3.0.2 if I attempt to execute on a canonicalized version of
CustomShuffleReaderExec I get an error "operating on canonicalized plan", as
expected.
There is a regression in Spark 3.1.1 where this check can never be reached
because of a new call to sendDriverMetrics that was added prior to the check.
This method will fail if operating on a canonicalized plan because it assumes
the existence of metrics that do not exist if this is a canonicalized plan.
{code:java}
private lazy val shuffleRDD: RDD[_] = {
sendDriverMetrics()
shuffleStage.map { stage =>
stage.shuffle.getShuffleRDD(partitionSpecs.toArray)
}.getOrElse {
throw new IllegalStateException("operating on canonicalized plan")
}
}{code}
The specific error looks like this:
{code:java}
java.util.NoSuchElementException: key not found: numPartitions
at scala.collection.immutable.Map$EmptyMap$.apply(Map.scala:101)
at scala.collection.immutable.Map$EmptyMap$.apply(Map.scala:99)
at
org.apache.spark.sql.execution.adaptive.CustomShuffleReaderExec.sendDriverMetrics(CustomShuffleReaderExec.scala:122)
at
org.apache.spark.sql.execution.adaptive.CustomShuffleReaderExec.shuffleRDD$lzycompute(CustomShuffleReaderExec.scala:182)
at
org.apache.spark.sql.execution.adaptive.CustomShuffleReaderExec.shuffleRDD(CustomShuffleReaderExec.scala:181)
at
org.apache.spark.sql.execution.adaptive.CustomShuffleReaderExec.doExecuteColumnar(CustomShuffleReaderExec.scala:196)
{code}
I think the fix is simply to avoid calling sendDriverMetrics if the plan is
canonicalized and I am planning on creating a PR to fix this.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
