[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16868178#comment-16868178
]
Josh Rosen commented on SPARK-26555:
Backported for 2.4.4 in
https://github.com/apache/spark/commit/ba7f61e25d58aa379f94a23b03503a25574529bc
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Assignee: Martin Loncaric
>Priority: Major
> Fix For: 2.4.4, 3.0.0
>
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16834203#comment-16834203
]
Josh Rosen commented on SPARK-26555:
I won't be able to tackle a backport for at least a week, so this is up for
grabs in case someone else wants to do it.
If I do end up working on this then I'll loop back here to claim it.
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Assignee: Martin Loncaric
>Priority: Major
> Fix For: 3.0.0
>
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16834055#comment-16834055
]
Sean Owen commented on SPARK-26555:
---
I personally think it's OK to backport -- do you want to open a PR and go for
it?
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Assignee: Martin Loncaric
>Priority: Major
> Fix For: 3.0.0
>
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16834025#comment-16834025
]
Josh Rosen commented on SPARK-26555:
[~cloud_fan] [~srowen], could we backport this to the 2.4.x series? It'd be
nice to have an LTS fix for users who can't immediately upgrade to 3.0.
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Assignee: Martin Loncaric
>Priority: Major
> Fix For: 3.0.0
>
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16792240#comment-16792240
]
Martin Loncaric commented on SPARK-26555:
-
This is an existing issue with scala: https://github.com/scala/bug/issues/10766
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16792086#comment-16792086
]
Martin Loncaric commented on SPARK-26555:
-
Update: I have been able to replicate this without Spark at all, using snippets
from org.apache.spark.sql.catalyst.ScalaReflection:
https://stackoverflow.com/questions/55150590/thread-safety-in-scala-reflection-with-type-matching
Investigating whether this can be fixed with different usage of the reflection
library, or whether this is a scala issue.
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16788773#comment-16788773
]
Martin Loncaric commented on SPARK-26555:
-
You can literally try any dataset and replicate this issue. For example,
sparkSession.createDataset(Seq(
MyClass(new Timestamp(1L), "b", "c", Some("d"), Some(1.0),
Some(2.0))
))
I think the code I left is pretty clear - it fails sometimes. Run it once, and
it may or may not work. I don't run multiple spark-submit's in parallel.
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16788771#comment-16788771
]
Sean Owen commented on SPARK-26555:
---
What is the fixed data set that reproduces this, to be clear?
And you mean that if you run it once it works, but fails in parallel?
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16782821#comment-16782821
]
Martin Loncaric commented on SPARK-26555:
-
Yes - when I take away any randomness and use the same dataset every time (say,
with Some(something) for each optional value), I still get this issue.
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16782806#comment-16782806
]
Sean Owen commented on SPARK-26555:
---
To be clear, is there a data set that works only when not run in this
concurrent loop? What I'm reading here is simply that you generate datasets
sometimes that (correctly, I believe) can't have their schema inferred.
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16782258#comment-16782258
]
Martin Loncaric commented on SPARK-26555:
-
I can also replicate with different schemas containing Option.
When I remove all Option columns from the schema, the sporadic failure goes
away. This also never happens when I remove the concurrency.
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16782239#comment-16782239
]
Martin Loncaric commented on SPARK-26555:
-
I was able to replicate with both all rows as `Some()` and all rows as `None`.
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16782185#comment-16782185
]
Martin Loncaric commented on SPARK-26555:
-
Will try it out and report back
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16782179#comment-16782179
]
Sean Owen commented on SPARK-26555:
---
Hm, I might have that backwards; might be when all are not None? at least, I
have a strong suspicious it's to do with the data that gets generated in some
runs. Maybe fix that at one data set and see if you can reproduce?
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16782146#comment-16782146
]
Sean Owen commented on SPARK-26555:
---
This doesn't look like a Spark bug. It comes up, I think, when your random data
set has all "None" for a column. That's what the error indicates at least. That
part of the code shouldn't have any shared state. Can you verify from your
debug output?
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-26555) Thread safety issue causes createDataset to fail with misleading errors
[
https://issues.apache.org/jira/browse/SPARK-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16735447#comment-16735447
]
Hyukjin Kwon commented on SPARK-26555:
--
Critical+ is usually reserved for committers. Please avoid to set this.
> Thread safety issue causes createDataset to fail with misleading errors
> ---
>
> Key: SPARK-26555
> URL: https://issues.apache.org/jira/browse/SPARK-26555
> Project: Spark
> Issue Type: Bug
> Components: SQL
>Affects Versions: 2.4.0
>Reporter: Martin Loncaric
>Priority: Major
>
> This can be replicated (~2% of the time) with
> {code:scala}
> import java.sql.Timestamp
> import java.util.concurrent.{Executors, Future}
> import org.apache.spark.sql.SparkSession
> import scala.collection.mutable.ListBuffer
> import scala.concurrent.ExecutionContext
> import scala.util.Random
> object Main {
> def main(args: Array[String]): Unit = {
> val sparkSession = SparkSession.builder
> .getOrCreate()
> import sparkSession.implicits._
> val executor = Executors.newFixedThreadPool(1)
> try {
> implicit val xc: ExecutionContext =
> ExecutionContext.fromExecutorService(executor)
> val futures = new ListBuffer[Future[_]]()
> for (i <- 1 to 3) {
> futures += executor.submit(new Runnable {
> override def run(): Unit = {
> val d = if (Random.nextInt(2) == 0) Some("d value") else None
> val e = if (Random.nextInt(2) == 0) Some(5.0) else None
> val f = if (Random.nextInt(2) == 0) Some(6.0) else None
> println("DEBUG", d, e, f)
> sparkSession.createDataset(Seq(
> MyClass(new Timestamp(1L), "b", "c", d, e, f)
> ))
> }
> })
> }
> futures.foreach(_.get())
> } finally {
> println("SHUTDOWN")
> executor.shutdown()
> sparkSession.stop()
> }
> }
> case class MyClass(
> a: Timestamp,
> b: String,
> c: String,
> d: Option[String],
> e: Option[Double],
> f: Option[Double]
> )
> }
> {code}
> So it will usually come up during
> {code:bash}
> for i in $(seq 1 200); do
> echo $i
> spark-submit --master local[4] target/scala-2.11/spark-test_2.11-0.1.jar
> done
> {code}
> causing a variety of possible errors, such as
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: scala.MatchError: scala.Option[String] (of class
> scala.reflect.internal.Types$ClassArgsTypeRef)
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$org$apache$spark$sql$catalyst$ScalaReflection$$deserializerFor$1.apply(ScalaReflection.scala:210){code}
> or
> {code}Exception in thread "main" java.util.concurrent.ExecutionException:
> java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at java.util.concurrent.FutureTask.report(FutureTask.java:122)
> Caused by: java.lang.UnsupportedOperationException: Schema for type
> scala.Option[scala.Double] is not supported
> at
> org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$schemaFor$1.apply(ScalaReflection.scala:789){code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
