[SPARK-3139] Akka timeouts from ContextCleaner when cleaning shuffles

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 1.1.0
• Fix Version/s: 1.1.0

Environment

10 r3.2xlarge tests on EC2, running the scala-agg-by-key-int spark-perf test against master commit d7e80c2597d4a9cae2e0cb35a86f7889323f4cbb.

• Target Version/s: 1.1.0
• Assignee: Guoqiang Li
• Reporter: Josh Rosen
• Created: 19/Aug/14 22:19
• Updated: 27/Aug/14 07:18
• Resolved: 27/Aug/14 07:18

Description

When running spark-perf tests on EC2, I have a job that's consistently logging the following Akka exceptions:

4/08/19 22:07:12 ERROR spark.ContextCleaner: Error cleaning shuffle 0
java.util.concurrent.TimeoutException: Futures timed out after [30 seconds]
  at scala.concurrent.impl.Promise$DefaultPromise.ready(Promise.scala:219)
  at scala.concurrent.impl.Promise$DefaultPromise.result(Promise.scala:223)
  at scala.concurrent.Await$$anonfun$result$1.apply(package.scala:107)
  at scala.concurrent.BlockContext$DefaultBlockContext$.blockOn(BlockContext.scala:53)
  at scala.concurrent.Await$.result(package.scala:107)
  at org.apache.spark.storage.BlockManagerMaster.removeShuffle(BlockManagerMaster.scala:118)
  at org.apache.spark.ContextCleaner.doCleanupShuffle(ContextCleaner.scala:159)
  at org.apache.spark.ContextCleaner$$anonfun$org$apache$spark$ContextCleaner$$keepCleaning$1$$anonfun$apply$mcV$sp$2.apply(ContextCleaner.scala:131)
  at org.apache.spark.ContextCleaner$$anonfun$org$apache$spark$ContextCleaner$$keepCleaning$1$$anonfun$apply$mcV$sp$2.apply(ContextCleaner.scala:124)
  at scala.Option.foreach(Option.scala:236)
  at org.apache.spark.ContextCleaner$$anonfun$org$apache$spark$ContextCleaner$$keepCleaning$1.apply$mcV$sp(ContextCleaner.scala:124)
  at org.apache.spark.ContextCleaner$$anonfun$org$apache$spark$ContextCleaner$$keepCleaning$1.apply(ContextCleaner.scala:120)
  at org.apache.spark.ContextCleaner$$anonfun$org$apache$spark$ContextCleaner$$keepCleaning$1.apply(ContextCleaner.scala:120)
  at org.apache.spark.util.Utils$.logUncaughtExceptions(Utils.scala:1252)
  at org.apache.spark.ContextCleaner.org$apache$spark$ContextCleaner$$keepCleaning(ContextCleaner.scala:119)
  at org.apache.spark.ContextCleaner$$anon$3.run(ContextCleaner.scala:65)

and

14/08/19 22:07:12 ERROR storage.BlockManagerMaster: Failed to remove shuffle 0
akka.pattern.AskTimeoutException: Timed out
  at akka.pattern.PromiseActorRef$$anonfun$1.apply$mcV$sp(AskSupport.scala:334)
  at akka.actor.Scheduler$$anon$11.run(Scheduler.scala:118)
  at scala.concurrent.Future$InternalCallbackExecutor$.scala$concurrent$Future$InternalCallbackExecutor$$unbatchedExecute(Future.scala:694)
  at scala.concurrent.Future$InternalCallbackExecutor$.execute(Future.scala:691)
  at akka.actor.LightArrayRevolverScheduler$TaskHolder.executeTask(Scheduler.scala:455)
  at akka.actor.LightArrayRevolverScheduler$$anon$12.executeBucket$1(Scheduler.scala:407)
  at akka.actor.LightArrayRevolverScheduler$$anon$12.nextTick(Scheduler.scala:411)
  at akka.actor.LightArrayRevolverScheduler$$anon$12.run(Scheduler.scala:363)
  at java.lang.Thread.run(Thread.java:745)

This doesn't seem to prevent the job from completing successfully, but it's serious issue because it means that resources aren't being cleaned up. The test script, ScalaAggByKeyInt, runs each test 10 times, and I see the same error after each test, so this seems deterministically reproducible.

I'll look at the executor logs to see if I can find more info there.

Issue Links

• is related to: SPARK-3015 Removing broadcast in quick successions causes Akka timeout
• links to: [Github] Pull Request #2056 (witgo)
• links to: [Github] Pull Request #2143 (tdas)