[SPARK-25297][Streaming][Test] Fix blocking unit tests for Scala 2.12

[da-liii]

What changes were proposed in this pull request?

Customize ExecutorContext's reporter to fix blocking unit tests for Scala 2.12

How was this patch tested?

./dev/change-scala-version.sh 2.12
$ sbt -Dscala-2.12
> ++2.12.6
> project streaming
> testOnly *FileBasedWriteAheadLogWithFileCloseAfterWriteSuite

[da-liii]

There is simplified project for this specific problem.

https://github.com/sadhen/blocking-future

[cloud-fan]

do you mean in scala 2.12 the default reporter is not throwing the exception?

[da-liii]

The default reporter does not throw the exception in Scala 2.11 either.

I guess there are bugs in the Future impl in Scala 2.12, since there is a big change at Future.scala in Scala 2.12.

This is a temporary fix.

I will dig into Future.scala, and hopefully propose a PR for Scala.

[srowen]

Thanks @sadhen I was just going to start looking into this. This seems like a reasonable solution.

[viktorklang]

The suggested solution will not work as intended as it will prevent subsequent callbacks (esp for other ExecutionContexts) to be submitted, esp. in the case of RejectedExecutionException.

See:
https://github.com/scala/scala/blob/2.12.x/src/library/scala/concurrent/impl/Promise.scala#L68
https://github.com/scala/scala/blob/2.12.x/src/library/scala/concurrent/impl/Promise.scala#L284

Or, in other words, rethrowing is not "reporting" the error.

The new Future encoding included in Scala 2.13.x deals much better with RejectedExecutionExceptions and InterruptedExceptions, but it is unclear right now what parts could be backported within the binary compatibility constraints.

[SparkQA]

Test build #95540 has finished for PR 22304 at commit 60eec77.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[srowen]

More info ... the test failure symptom is that this test never completes, just repeating a RejectedExecutionException:

- FileBasedWriteAheadLog - handling file errors while reading rotating logs
- FileBasedWriteAheadLog - do not create directories or files unless write
java.util.concurrent.RejectedExecutionException
	at java.util.concurrent.ForkJoinPool.externalSubmit(ForkJoinPool.java:2328)
	at java.util.concurrent.ForkJoinPool.externalPush(ForkJoinPool.java:2419)
	at java.util.concurrent.ForkJoinPool.execute(ForkJoinPool.java:2648)
	at scala.concurrent.impl.ExecutionContextImpl$$anon$1.execute(ExecutionContextImpl.scala:134)
	at scala.concurrent.impl.CallbackRunnable.executeWithValue(Promise.scala:68)
	at scala.concurrent.impl.Promise$KeptPromise$Kept.onComplete(Promise.scala:368)
	at scala.concurrent.impl.Promise$KeptPromise$Kept.onComplete$(Promise.scala:367)
	at scala.concurrent.impl.Promise$KeptPromise$Successful.onComplete(Promise.scala:375)
	at scala.concurrent.impl.Promise.transform(Promise.scala:29)
	at scala.concurrent.impl.Promise.transform$(Promise.scala:27)
	at scala.concurrent.impl.Promise$KeptPromise$Successful.transform(Promise.scala:375)
	at scala.concurrent.Future.map(Future.scala:288)
	at scala.concurrent.Future.map$(Future.scala:288)
	at scala.concurrent.impl.Promise$KeptPromise$Successful.map(Promise.scala:375)
	at scala.concurrent.Future$.apply(Future.scala:654)
	at org.apache.spark.util.ThreadUtils$.$anonfun$parmap$1(ThreadUtils.scala:314)
	at scala.collection.TraversableLike.$anonfun$map$1(TraversableLike.scala:234)
	at scala.collection.mutable.ResizableArray.foreach(ResizableArray.scala:59)
	at scala.collection.mutable.ResizableArray.foreach$(ResizableArray.scala:52)
	at scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:48)
	at scala.collection.TraversableLike.map(TraversableLike.scala:234)
	at scala.collection.TraversableLike.map$(TraversableLike.scala:227)
	at scala.collection.AbstractTraversable.map(Traversable.scala:104)
	at org.apache.spark.util.ThreadUtils$.parmap(ThreadUtils.scala:314)
	at org.apache.spark.streaming.util.FileBasedWriteAheadLog$.$anonfun$seqToParIterator$1(FileBasedWriteAheadLog.scala:318)
	at scala.collection.Iterator$$anon$11.nextCur(Iterator.scala:482)
	at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:488)
	at scala.collection.Iterator$$anon$10.hasNext(Iterator.scala:456)
	at scala.collection.TraversableOnce$FlattenOps$$anon$1.hasNext(TraversableOnce.scala:464)
	at scala.collection.Iterator$$anon$10.hasNext(Iterator.scala:456)
	at scala.collection.Iterator.foreach(Iterator.scala:944)
	at scala.collection.Iterator.foreach$(Iterator.scala:944)
	at scala.collection.AbstractIterator.foreach(Iterator.scala:1432)
	at scala.collection.generic.Growable.$plus$plus$eq(Growable.scala:59)
	at scala.collection.generic.Growable.$plus$plus$eq$(Growable.scala:50)
	at scala.collection.mutable.ArrayBuffer.$plus$plus$eq(ArrayBuffer.scala:104)
	at scala.collection.mutable.ArrayBuffer.$plus$plus$eq(ArrayBuffer.scala:48)
	at scala.collection.TraversableOnce.to(TraversableOnce.scala:310)
	at scala.collection.TraversableOnce.to$(TraversableOnce.scala:308)
	at scala.collection.AbstractIterator.to(Iterator.scala:1432)
	at scala.collection.TraversableOnce.toBuffer(TraversableOnce.scala:302)
	at scala.collection.TraversableOnce.toBuffer$(TraversableOnce.scala:302)
	at scala.collection.AbstractIterator.toBuffer(Iterator.scala:1432)
	at scala.collection.TraversableOnce.toArray(TraversableOnce.scala:289)
	at scala.collection.TraversableOnce.toArray$(TraversableOnce.scala:283)
	at scala.collection.AbstractIterator.toArray(Iterator.scala:1432)
	at org.apache.spark.streaming.util.CommonWriteAheadLogTests.$anonfun$new$16(WriteAheadLogSuite.scala:213)
	at org.scalatest.Assertions.intercept(Assertions.scala:805)
	at org.scalatest.Assertions.intercept$(Assertions.scala:802)
	at org.scalatest.FunSuite.intercept(FunSuite.scala:1560)
	at org.apache.spark.streaming.util.CommonWriteAheadLogTests.$anonfun$new$13(WriteAheadLogSuite.scala:213)

[srowen]

The one concern is whether this behavior change impacts anything else. Now some exceptions will cause an actual exception in a caller to methods like Await, rather than just logging and continuing. That might even be a good thing, and is crucial here. But that's what I have in mind. That said, tests all pass.

[cloud-fan]

@srowen are we targeting this patch to Spark 2.4?

[da-liii]

@srowen @cloud-fan The merged #22292 already fixed it.