[jira] [Commented] (FLINK-7279) MiniCluster can deadlock at shut down
[ https://issues.apache.org/jira/browse/FLINK-7279?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16106586#comment-16106586 ] ASF GitHub Bot commented on FLINK-7279: --- Github user asfgit closed the pull request at: https://github.com/apache/flink/pull/4416 > MiniCluster can deadlock at shut down > - > > Key: FLINK-7279 > URL: https://issues.apache.org/jira/browse/FLINK-7279 > Project: Flink > Issue Type: Bug > Components: Tests >Affects Versions: 1.4.0 >Reporter: Till Rohrmann >Assignee: Nico Kruber > Labels: flip-6 > Fix For: 1.4.0 > > > The {{MiniCluster}} can deadlock in case if the fatal error handler is called > while the {{MiniCluster}} shuts down. The reason is that the shut down > happens under a lock which is required by the fatal error handler as well. If > now the {{MiniCluster}} tries to shut down the underlying RPC service which > waits for all actors to terminate, it will never complete because one actor > is still waiting for the lock. > One solution would be to ignore the fatal error handler calls if the > {{MiniCluster}} is shutting down. > https://s3.amazonaws.com/archive.travis-ci.org/jobs/257811319/log.txt -- This message was sent by Atlassian JIRA (v6.4.14#64029)
[jira] [Commented] (FLINK-7279) MiniCluster can deadlock at shut down
[ https://issues.apache.org/jira/browse/FLINK-7279?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16105146#comment-16105146 ] ASF GitHub Bot commented on FLINK-7279: --- Github user NicoK commented on the issue: https://github.com/apache/flink/pull/4416 ok, I think, I understood the intention of `TaskExecutor#onFatalErrorAsync` wrong - it is meant for outside `TaskExecutor` thread calls to run the error handler inside the `TaskExecutor` thread... I'll create a new approach > MiniCluster can deadlock at shut down > - > > Key: FLINK-7279 > URL: https://issues.apache.org/jira/browse/FLINK-7279 > Project: Flink > Issue Type: Bug > Components: Tests >Affects Versions: 1.4.0 >Reporter: Till Rohrmann > Labels: flip-6 > > The {{MiniCluster}} can deadlock in case if the fatal error handler is called > while the {{MiniCluster}} shuts down. The reason is that the shut down > happens under a lock which is required by the fatal error handler as well. If > now the {{MiniCluster}} tries to shut down the underlying RPC service which > waits for all actors to terminate, it will never complete because one actor > is still waiting for the lock. > One solution would be to ignore the fatal error handler calls if the > {{MiniCluster}} is shutting down. > https://s3.amazonaws.com/archive.travis-ci.org/jobs/257811319/log.txt -- This message was sent by Atlassian JIRA (v6.4.14#64029)
[jira] [Commented] (FLINK-7279) MiniCluster can deadlock at shut down
[ https://issues.apache.org/jira/browse/FLINK-7279?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16105069#comment-16105069 ] ASF GitHub Bot commented on FLINK-7279: --- Github user NicoK commented on the issue: https://github.com/apache/flink/pull/4416 A simple asynchronous call as in `TaskExecutor#onFatalErrorAsync()` is not enough though because this is what is actually already done and led to me finding this error in the first place. Please see the stack traces of the incriminating processes below: ``` "flink-akka.actor.default-dispatcher-8" #31 prio=5 os_prio=0 tid=0x7ffa00efe800 nid=0xb8c waiting for monitor entry [0x7ff9ee54] java.lang.Thread.State: BLOCKED (on object monitor) at org.apache.flink.runtime.minicluster.MiniCluster$TerminatingFatalErrorHandler.onFatalError(MiniCluster.java:652) - waiting to lock <0xaad1d2d8> (a java.lang.Object) at org.apache.flink.runtime.taskexecutor.TaskExecutor.onFatalError(TaskExecutor.java:1129) at org.apache.flink.runtime.taskexecutor.TaskExecutor$7.run(TaskExecutor.java:1116) at org.apache.flink.runtime.rpc.akka.AkkaRpcActor.handleRunAsync(AkkaRpcActor.java:278) at org.apache.flink.runtime.rpc.akka.AkkaRpcActor.handleMessage(AkkaRpcActor.java:132) at org.apache.flink.runtime.rpc.akka.AkkaRpcActor.access$000(AkkaRpcActor.java:73) at org.apache.flink.runtime.rpc.akka.AkkaRpcActor$1.apply(AkkaRpcActor.java:111) at akka.actor.ActorCell$$anonfun$become$1.applyOrElse(ActorCell.scala:534) at akka.actor.Actor$class.aroundReceive(Actor.scala:467) at akka.actor.UntypedActor.aroundReceive(UntypedActor.scala:97) at akka.actor.ActorCell.receiveMessage(ActorCell.scala:516) at akka.actor.ActorCell.invoke(ActorCell.scala:487) at akka.dispatch.Mailbox.processMailbox(Mailbox.scala:238) at akka.dispatch.Mailbox.run(Mailbox.scala:220) at akka.dispatch.ForkJoinExecutorConfigurator$AkkaForkJoinTask.exec(AbstractDispatcher.scala:397) at scala.concurrent.forkjoin.ForkJoinTask.doExec(ForkJoinTask.java:260) at scala.concurrent.forkjoin.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:1339) at scala.concurrent.forkjoin.ForkJoinPool.runWorker(ForkJoinPool.java:1979) at scala.concurrent.forkjoin.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:107) "main" #1 prio=5 os_prio=0 tid=0x7ffaa000 nid=0xb56 waiting on condition [0x7ffa07fa1000] java.lang.Thread.State: WAITING (parking) at sun.misc.Unsafe.park(Native Method) - parking to wait for <0xab269c30> (a java.util.concurrent.CountDownLatch$Sync) at java.util.concurrent.locks.LockSupport.park(LockSupport.java:175) at java.util.concurrent.locks.AbstractQueuedSynchronizer.parkAndCheckInterrupt(AbstractQueuedSynchronizer.java:836) at java.util.concurrent.locks.AbstractQueuedSynchronizer.doAcquireSharedInterruptibly(AbstractQueuedSynchronizer.java:997) at java.util.concurrent.locks.AbstractQueuedSynchronizer.acquireSharedInterruptibly(AbstractQueuedSynchronizer.java:1304) at java.util.concurrent.CountDownLatch.await(CountDownLatch.java:231) at akka.actor.ActorSystemImpl$TerminationCallbacks.ready(ActorSystem.scala:819) at akka.actor.ActorSystemImpl$TerminationCallbacks.ready(ActorSystem.scala:788) at scala.concurrent.Await$$anonfun$ready$1.apply(package.scala:169) at scala.concurrent.Await$$anonfun$ready$1.apply(package.scala:169) at scala.concurrent.BlockContext$DefaultBlockContext$.blockOn(BlockContext.scala:53) at scala.concurrent.Await$.ready(package.scala:169) at akka.actor.ActorSystemImpl.awaitTermination(ActorSystem.scala:644) at akka.actor.ActorSystemImpl.awaitTermination(ActorSystem.scala:645) at org.apache.flink.runtime.rpc.akka.AkkaRpcService.stopService(AkkaRpcService.java:282) at org.apache.flink.runtime.minicluster.MiniCluster.shutDownRpc(MiniCluster.java:596) at org.apache.flink.runtime.minicluster.MiniCluster.shutdownInternally(MiniCluster.java:364) at org.apache.flink.runtime.minicluster.MiniCluster.shutdown(MiniCluster.java:309) - locked <0xaad1d2d8> (a java.lang.Object) at org.apache.flink.runtime.minicluster.MiniClusterITCase.runJobWithMultipleJobManagers(MiniClusterITCase.java:87) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:498) at org.junit.runners.model.FrameworkMethod$1.runReflectiveCall(FrameworkMethod.java:50)
[jira] [Commented] (FLINK-7279) MiniCluster can deadlock at shut down
[ https://issues.apache.org/jira/browse/FLINK-7279?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16104746#comment-16104746 ] ASF GitHub Bot commented on FLINK-7279: --- GitHub user NicoK opened a pull request: https://github.com/apache/flink/pull/4416 [FLINK-7279][minicluster] fix a deadlock between TM and cluster shutdown ## What is the purpose of the change The `MiniCluster` can deadlock if the fatal error handler is called while the `MiniCluster` shuts down. The reason is that the shut down happens under a lock which is required by the fatal error handler as well. If now the `MiniCluster` tries to shut down the underlying RPC service which waits for all actors to terminate, it will never complete because one actor is still waiting for the lock. ## Brief change log - guard both shutdown methods by a new `ReentrantLock` and ignore the TM shutdown in the `TerminatingFatalErrorHandler` if the cluster is already shut down. ## Verifying this change This change is already covered by existing tests, such as `MiniClusterITCase` which was instable because of this bug (also see [FLINK-7115]). ## Does this pull request potentially affect one of the following parts: - Dependencies (does it add or upgrade a dependency): (no) - The public API, i.e., is any changed class annotated with `@Public(Evolving)`: (no) - The serializers: (no) - The runtime per-record code paths (performance sensitive): (no) - Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Yarn/Mesos, ZooKeeper: (no) ## Documentation - Does this pull request introduce a new feature? (no) - If yes, how is the feature documented? (JavaDocs) You can merge this pull request into a Git repository by running: $ git pull https://github.com/NicoK/flink flink-7279 Alternatively you can review and apply these changes as the patch at: https://github.com/apache/flink/pull/4416.patch To close this pull request, make a commit to your master/trunk branch with (at least) the following in the commit message: This closes #4416 commit 7815deb974316d58da0646bd26ce718bbd597ba7 Author: Nico Kruber Date: 2017-07-28T09:45:48Z [FLINK-7279][minicluster] fix a deadlock between TM and cluster shutdown The MiniCluster can deadlock if the fatal error handler is called while the MiniCluster shuts down. The reason is that the shut down happens under a lock which is required by the fatal error handler as well. If now the MiniCluster tries to shut down the underlying RPC service which waits for all actors to terminate, it will never complete because one actor is still waiting for the lock. Solution: guard both shutdown methods by a new ReentrantLock and ignore the TM shutdown in the TerminatingFatalErrorHandler if the cluster is already shut down. > MiniCluster can deadlock at shut down > - > > Key: FLINK-7279 > URL: https://issues.apache.org/jira/browse/FLINK-7279 > Project: Flink > Issue Type: Bug > Components: Tests >Affects Versions: 1.4.0 >Reporter: Till Rohrmann > Labels: flip-6 > > The {{MiniCluster}} can deadlock in case if the fatal error handler is called > while the {{MiniCluster}} shuts down. The reason is that the shut down > happens under a lock which is required by the fatal error handler as well. If > now the {{MiniCluster}} tries to shut down the underlying RPC service which > waits for all actors to terminate, it will never complete because one actor > is still waiting for the lock. > One solution would be to ignore the fatal error handler calls if the > {{MiniCluster}} is shutting down. > https://s3.amazonaws.com/archive.travis-ci.org/jobs/257811319/log.txt -- This message was sent by Atlassian JIRA (v6.4.14#64029)