[jira] [Comment Edited] (FLINK-29234) Dead lock in DefaultLeaderElectionService
[
https://issues.apache.org/jira/browse/FLINK-29234?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17627706#comment-17627706
]
Weijie Guo edited comment on FLINK-29234 at 11/2/22 1:35 PM:
-
[~fpaul] , The PR changes the production code very little, and it may take 1～2
days to complete the review.
was (Author: weijie guo):
[~fpaul] , The PR changes the production code very little, and it may take 1～2
days to complete the review
> Dead lock in DefaultLeaderElectionService
> -
>
> Key: FLINK-29234
> URL: https://issues.apache.org/jira/browse/FLINK-29234
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Coordination
>Affects Versions: 1.13.5, 1.14.5, 1.15.2
>Reporter: Yu Wang
>Assignee: Weijie Guo
>Priority: Critical
> Labels: pull-request-available
>
> Jobmanager stop working because the deadlock in DefaultLeaderElectionService.
> The log stopped at
> {code:java}
> org.apache.flink.runtime.leaderelection.DefaultLeaderElectionService [] -
> Stopping DefaultLeaderElectionService. {code}
> Which may similar to this ticket
> https://issues.apache.org/jira/browse/FLINK-20008
> Here is the jstack info
> {code:java}
> Found one Java-level deadlock:
> =
> "flink-akka.actor.default-dispatcher-18": waiting to lock monitor
> 0x7f15c7eae3a8 (object 0x000678d395e8, a java.lang.Object), which is
> held by "main-EventThread" "main-EventThread": waiting to lock monitor
> 0x7f15a3811258 (object 0x000678cf1be0, a java.lang.Object), which is
> held by "flink-akka.actor.default-dispatcher-18" Java stack information for
> the threads listed above:
> ===
> "flink-akka.actor.default-dispatcher-18":
> at
> org.apache.flink.runtime.leaderelection.DefaultLeaderElectionService.stop(DefaultLeaderElectionService.java:104)
> - waiting to lock <0x000678d395e8> (a java.lang.Object)
> at
> org.apache.flink.runtime.jobmaster.JobMasterServiceLeadershipRunner.lambda$closeAsync$0(JobMasterServiceLeadershipRunner.java:147)
> at
> org.apache.flink.runtime.jobmaster.JobMasterServiceLeadershipRunner$$Lambda$735/1742012752.run(Unknown
> Source)
> at
> org.apache.flink.runtime.concurrent.FutureUtils.lambda$runAfterwardsAsync$18(FutureUtils.java:687)
> at
> org.apache.flink.runtime.concurrent.FutureUtils$$Lambda$736/6716561.accept(Unknown
> Source)
> at
> java.util.concurrent.CompletableFuture.uniWhenComplete(CompletableFuture.java:774)
> at
> java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(CompletableFuture.java:750)
> at
> java.util.concurrent.CompletableFuture$Completion.run(CompletableFuture.java:456)
> at
> org.apache.flink.runtime.concurrent.DirectExecutorService.execute(DirectExecutorService.java:217)
> at
> java.util.concurrent.CompletableFuture$UniCompletion.claim(CompletableFuture.java:543)
> at
> java.util.concurrent.CompletableFuture.uniWhenComplete(CompletableFuture.java:765)
> at
> java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(CompletableFuture.java:750)
> at
> java.util.concurrent.CompletableFuture.uniWhenCompleteStage(CompletableFuture.java:795)
> at
> java.util.concurrent.CompletableFuture.whenCompleteAsync(CompletableFuture.java:2163)
> at
> org.apache.flink.runtime.concurrent.FutureUtils.runAfterwardsAsync(FutureUtils.java:684)
> at
> org.apache.flink.runtime.concurrent.FutureUtils.runAfterwards(FutureUtils.java:651)
> at
> org.apache.flink.runtime.jobmaster.JobMasterServiceLeadershipRunner.closeAsync(JobMasterServiceLeadershipRunner.java:143)
> - locked <0x000678cf1be0> (a java.lang.Object)
> at
> org.apache.flink.runtime.dispatcher.Dispatcher.terminateJob(Dispatcher.java:807)
> at
> org.apache.flink.runtime.dispatcher.Dispatcher.terminateRunningJobs(Dispatcher.java:799)
> at
> org.apache.flink.runtime.dispatcher.Dispatcher.terminateRunningJobsAndGetTerminationFuture(Dispatcher.java:812)
> at org.apache.flink.runtime.dispatcher.Dispatcher.onStop(Dispatcher.java:268)
> at
> org.apache.flink.runtime.rpc.RpcEndpoint.internalCallOnStop(RpcEndpoint.java:214)
> at
> org.apache.flink.runtime.rpc.akka.AkkaRpcActor$StartedState.terminate(AkkaRpcActor.java:563)
> at
> org.apache.flink.runtime.rpc.akka.AkkaRpcActor.handleControlMessage(AkkaRpcActor.java:186)
> at
> org.apache.flink.runtime.rpc.akka.AkkaRpcActor$$Lambda$444/1289054037.apply(Unknown
> Source)
> at akka.japi.pf.UnitCaseStatement.apply(CaseStatements.scala:26)
> at akka.japi.pf.UnitCaseStatement.apply(CaseStatements.scala:21)
> at scala.PartialFunction.applyOrElse(PartialFunction.scala:123)
> at scala.PartialFunction.applyOrElse$(PartialFunction.scala:122)
> at
[jira] [Comment Edited] (FLINK-29234) Dead lock in DefaultLeaderElectionService
[
https://issues.apache.org/jira/browse/FLINK-29234?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17627706#comment-17627706
]
Weijie Guo edited comment on FLINK-29234 at 11/2/22 1:30 PM:
-
[~fpaul] , The PR changes the production code very little, and it may take 1～2
days to complete the review
was (Author: weijie guo):
[~fpaul] , The PR changes the production code very little, and it may take 1～2
days to review.
> Dead lock in DefaultLeaderElectionService
> -
>
> Key: FLINK-29234
> URL: https://issues.apache.org/jira/browse/FLINK-29234
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Coordination
>Affects Versions: 1.13.5, 1.14.5, 1.15.2
>Reporter: Yu Wang
>Assignee: Weijie Guo
>Priority: Critical
> Labels: pull-request-available
>
> Jobmanager stop working because the deadlock in DefaultLeaderElectionService.
> The log stopped at
> {code:java}
> org.apache.flink.runtime.leaderelection.DefaultLeaderElectionService [] -
> Stopping DefaultLeaderElectionService. {code}
> Which may similar to this ticket
> https://issues.apache.org/jira/browse/FLINK-20008
> Here is the jstack info
> {code:java}
> Found one Java-level deadlock:
> =
> "flink-akka.actor.default-dispatcher-18": waiting to lock monitor
> 0x7f15c7eae3a8 (object 0x000678d395e8, a java.lang.Object), which is
> held by "main-EventThread" "main-EventThread": waiting to lock monitor
> 0x7f15a3811258 (object 0x000678cf1be0, a java.lang.Object), which is
> held by "flink-akka.actor.default-dispatcher-18" Java stack information for
> the threads listed above:
> ===
> "flink-akka.actor.default-dispatcher-18":
> at
> org.apache.flink.runtime.leaderelection.DefaultLeaderElectionService.stop(DefaultLeaderElectionService.java:104)
> - waiting to lock <0x000678d395e8> (a java.lang.Object)
> at
> org.apache.flink.runtime.jobmaster.JobMasterServiceLeadershipRunner.lambda$closeAsync$0(JobMasterServiceLeadershipRunner.java:147)
> at
> org.apache.flink.runtime.jobmaster.JobMasterServiceLeadershipRunner$$Lambda$735/1742012752.run(Unknown
> Source)
> at
> org.apache.flink.runtime.concurrent.FutureUtils.lambda$runAfterwardsAsync$18(FutureUtils.java:687)
> at
> org.apache.flink.runtime.concurrent.FutureUtils$$Lambda$736/6716561.accept(Unknown
> Source)
> at
> java.util.concurrent.CompletableFuture.uniWhenComplete(CompletableFuture.java:774)
> at
> java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(CompletableFuture.java:750)
> at
> java.util.concurrent.CompletableFuture$Completion.run(CompletableFuture.java:456)
> at
> org.apache.flink.runtime.concurrent.DirectExecutorService.execute(DirectExecutorService.java:217)
> at
> java.util.concurrent.CompletableFuture$UniCompletion.claim(CompletableFuture.java:543)
> at
> java.util.concurrent.CompletableFuture.uniWhenComplete(CompletableFuture.java:765)
> at
> java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(CompletableFuture.java:750)
> at
> java.util.concurrent.CompletableFuture.uniWhenCompleteStage(CompletableFuture.java:795)
> at
> java.util.concurrent.CompletableFuture.whenCompleteAsync(CompletableFuture.java:2163)
> at
> org.apache.flink.runtime.concurrent.FutureUtils.runAfterwardsAsync(FutureUtils.java:684)
> at
> org.apache.flink.runtime.concurrent.FutureUtils.runAfterwards(FutureUtils.java:651)
> at
> org.apache.flink.runtime.jobmaster.JobMasterServiceLeadershipRunner.closeAsync(JobMasterServiceLeadershipRunner.java:143)
> - locked <0x000678cf1be0> (a java.lang.Object)
> at
> org.apache.flink.runtime.dispatcher.Dispatcher.terminateJob(Dispatcher.java:807)
> at
> org.apache.flink.runtime.dispatcher.Dispatcher.terminateRunningJobs(Dispatcher.java:799)
> at
> org.apache.flink.runtime.dispatcher.Dispatcher.terminateRunningJobsAndGetTerminationFuture(Dispatcher.java:812)
> at org.apache.flink.runtime.dispatcher.Dispatcher.onStop(Dispatcher.java:268)
> at
> org.apache.flink.runtime.rpc.RpcEndpoint.internalCallOnStop(RpcEndpoint.java:214)
> at
> org.apache.flink.runtime.rpc.akka.AkkaRpcActor$StartedState.terminate(AkkaRpcActor.java:563)
> at
> org.apache.flink.runtime.rpc.akka.AkkaRpcActor.handleControlMessage(AkkaRpcActor.java:186)
> at
> org.apache.flink.runtime.rpc.akka.AkkaRpcActor$$Lambda$444/1289054037.apply(Unknown
> Source)
> at akka.japi.pf.UnitCaseStatement.apply(CaseStatements.scala:26)
> at akka.japi.pf.UnitCaseStatement.apply(CaseStatements.scala:21)
> at scala.PartialFunction.applyOrElse(PartialFunction.scala:123)
> at scala.PartialFunction.applyOrElse$(PartialFunction.scala:122)
> at akka.japi.pf.UnitCaseStatement.applyOrElse(CaseStatements.scala:21)
> at
[jira] [Comment Edited] (FLINK-29234) Dead lock in DefaultLeaderElectionService
[
https://issues.apache.org/jira/browse/FLINK-29234?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17602689#comment-17602689
]
Yu Wang edited comment on FLINK-29234 at 9/10/22 9:48 AM:
--
[~martijnvisser] I think this issue still exists in the master branch. The
issue is caused by the ordering of Flink to get the lock.
In this line, the class *DefaultLeaderElectionService* will try to get the lock
of itself, then invoke the method of
*leaderContender(which is JobMasterServiceLeaderShipRunner in this case).*
[https://github.com/apache/flink/blob/master/flink-runtime/src/main/java/org/apache/flink/runtime/leaderelection/DefaultLeaderElectionService.java#L204]
So the order of getting locks is
# DefaultLeaderElectionService
# JobMasterServiceLeaderShipRunner
And in this line *JobMasterServiceLeaderShipRunner* will try to get the lock of
itself, then invoke the method of *leaderElectionService(which is*
{*}DefaultLeaderElectionService in this case{*}{*}).{*}
[https://github.com/apache/flink/blob/master/flink-runtime/src/main/java/org/apache/flink/runtime/jobmaster/JobMasterServiceLeadershipRunner.java#L148]
So the order of getting the locks is
# JobMasterServiceLeaderShipRunner
# DefaultLeaderElectionService
So if these two functions are invoked nearly at the same time, it will cause
the dead lock issue.
was (Author: lucentwong):
[~martijnvisser] I think this issue still exists in the master branch. The
issue is caused by the ordering of Flink to get the lock.
In this line, the class *DefaultLeaderElectionService* will try to get the lock
of itself, then invoke the method of
*leaderContender(which is JobMasterServiceLeaderShipRunner in this case).*
[https://github.com/apache/flink/blob/master/flink-runtime/src/main/java/org/apache/flink/runtime/leaderelection/DefaultLeaderElectionService.java#L204]
So the order of getting locks is
# DefaultLeaderElectionService
# JobMasterServiceLeaderShipRunner{*}{*}
And in this line *JobMasterServiceLeaderShipRunner* will try to get the lock of
itself, then invoke the method of *leaderElectionService(which is*
{*}DefaultLeaderElectionService in this case{*}{*}).{*}
[https://github.com/apache/flink/blob/master/flink-runtime/src/main/java/org/apache/flink/runtime/jobmaster/JobMasterServiceLeadershipRunner.java#L148]
So the order of getting the locks is
# JobMasterServiceLeaderShipRunner
# DefaultLeaderElectionService
So if these two functions are invoked nearly at the same time, it will cause
the dead lock issue.
> Dead lock in DefaultLeaderElectionService
> -
>
> Key: FLINK-29234
> URL: https://issues.apache.org/jira/browse/FLINK-29234
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Coordination
>Affects Versions: 1.13.5
>Reporter: Yu Wang
>Priority: Critical
>
> Jobmanager stop working because the deadlock in DefaultLeaderElectionService.
> The log stopped at
> {code:java}
> org.apache.flink.runtime.leaderelection.DefaultLeaderElectionService [] -
> Stopping DefaultLeaderElectionService. {code}
> Which may similar to this ticket
> https://issues.apache.org/jira/browse/FLINK-20008
> Here is the jstack info
> {code:java}
> Found one Java-level deadlock: =
> "flink-akka.actor.default-dispatcher-18": waiting to lock monitor
> 0x7f15c7eae3a8 (object 0x000678d395e8, a java.lang.Object), which is
> held by "main-EventThread" "main-EventThread": waiting to lock monitor
> 0x7f15a3811258 (object 0x000678cf1be0, a java.lang.Object), which is
> held by "flink-akka.actor.default-dispatcher-18" Java stack information for
> the threads listed above: ===
> "flink-akka.actor.default-dispatcher-18": at
> org.apache.flink.runtime.leaderelection.DefaultLeaderElectionService.stop(DefaultLeaderElectionService.java:104)
> - waiting to lock <0x000678d395e8> (a java.lang.Object) at
> org.apache.flink.runtime.jobmaster.JobMasterServiceLeadershipRunner.lambda$closeAsync$0(JobMasterServiceLeadershipRunner.java:147)
> at
> org.apache.flink.runtime.jobmaster.JobMasterServiceLeadershipRunner$$Lambda$735/1742012752.run(Unknown
> Source) at
> org.apache.flink.runtime.concurrent.FutureUtils.lambda$runAfterwardsAsync$18(FutureUtils.java:687)
> at
> org.apache.flink.runtime.concurrent.FutureUtils$$Lambda$736/6716561.accept(Unknown
> Source) at
> java.util.concurrent.CompletableFuture.uniWhenComplete(CompletableFuture.java:774)
> at
> java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(CompletableFuture.java:750)
> at
> java.util.concurrent.CompletableFuture$Completion.run(CompletableFuture.java:456)
> at
>
[jira] [Comment Edited] (FLINK-29234) Dead lock in DefaultLeaderElectionService
[
https://issues.apache.org/jira/browse/FLINK-29234?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17602689#comment-17602689
]
Yu Wang edited comment on FLINK-29234 at 9/10/22 9:47 AM:
--
[~martijnvisser] I think this issue still exists in the master branch. The
issue is caused by the ordering of Flink to get the lock.
In this line, the class *DefaultLeaderElectionService* will try to get the lock
of itself, then invoke the method of
*leaderContender(which is JobMasterServiceLeaderShipRunner in this case).*
[https://github.com/apache/flink/blob/master/flink-runtime/src/main/java/org/apache/flink/runtime/leaderelection/DefaultLeaderElectionService.java#L204]
So the order of getting locks is
# DefaultLeaderElectionService
# JobMasterServiceLeaderShipRunner{*}{*}
And in this line *JobMasterServiceLeaderShipRunner* will try to get the lock of
itself, then invoke the method of *leaderElectionService(which is*
{*}DefaultLeaderElectionService in this case{*}{*}).{*}
[https://github.com/apache/flink/blob/master/flink-runtime/src/main/java/org/apache/flink/runtime/jobmaster/JobMasterServiceLeadershipRunner.java#L148]
So the order of getting the locks is
# JobMasterServiceLeaderShipRunner
# DefaultLeaderElectionService
So if these two functions are invoked nearly at the same time, it will cause
the dead lock issue.
was (Author: lucentwong):
[~martijnvisser] I think this issue still exists in the master branch. The
issue is caused by the ordering of Flink to get the lock.
In this line, the class *DefaultLeaderElectionService* will try to get the lock
of itself, then invoke the method of
*leaderContender(which is JobMasterServiceLeaderShipRunner in this case).*
[https://github.com/apache/flink/blob/master/flink-runtime/src/main/java/org/apache/flink/runtime/leaderelection/DefaultLeaderElectionService.java#L204]
So the order of getting locks is
# DefaultLeaderElectionService
# JobMasterServiceLeaderShipRunner{*}{*}
And in this line *JobMasterServiceLeaderShipRunner* will try to get the lock of
itself, then invoke the method of *leaderElectionService(which is*
{*}DefaultLeaderElectionService in this case{*}{*}).{*}
[https://github.com/apache/flink/blob/master/flink-runtime/src/main/java/org/apache/flink/runtime/jobmaster/JobMasterServiceLeadershipRunner.java#L148]
So the order of getting the locks is
# JobMasterServiceLeaderShipRunner
# DefaultLeaderElectionService
So if these two functions are invoked nearly at the same time, it will cause
the dead lock issue.
> Dead lock in DefaultLeaderElectionService
> -
>
> Key: FLINK-29234
> URL: https://issues.apache.org/jira/browse/FLINK-29234
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Coordination
>Affects Versions: 1.13.5
>Reporter: Yu Wang
>Priority: Critical
>
> Jobmanager stop working because the deadlock in DefaultLeaderElectionService.
> The log stopped at
> {code:java}
> org.apache.flink.runtime.leaderelection.DefaultLeaderElectionService [] -
> Stopping DefaultLeaderElectionService. {code}
> Which may similar to this ticket
> https://issues.apache.org/jira/browse/FLINK-20008
> Here is the jstack info
> {code:java}
> Found one Java-level deadlock: =
> "flink-akka.actor.default-dispatcher-18": waiting to lock monitor
> 0x7f15c7eae3a8 (object 0x000678d395e8, a java.lang.Object), which is
> held by "main-EventThread" "main-EventThread": waiting to lock monitor
> 0x7f15a3811258 (object 0x000678cf1be0, a java.lang.Object), which is
> held by "flink-akka.actor.default-dispatcher-18" Java stack information for
> the threads listed above: ===
> "flink-akka.actor.default-dispatcher-18": at
> org.apache.flink.runtime.leaderelection.DefaultLeaderElectionService.stop(DefaultLeaderElectionService.java:104)
> - waiting to lock <0x000678d395e8> (a java.lang.Object) at
> org.apache.flink.runtime.jobmaster.JobMasterServiceLeadershipRunner.lambda$closeAsync$0(JobMasterServiceLeadershipRunner.java:147)
> at
> org.apache.flink.runtime.jobmaster.JobMasterServiceLeadershipRunner$$Lambda$735/1742012752.run(Unknown
> Source) at
> org.apache.flink.runtime.concurrent.FutureUtils.lambda$runAfterwardsAsync$18(FutureUtils.java:687)
> at
> org.apache.flink.runtime.concurrent.FutureUtils$$Lambda$736/6716561.accept(Unknown
> Source) at
> java.util.concurrent.CompletableFuture.uniWhenComplete(CompletableFuture.java:774)
> at
> java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(CompletableFuture.java:750)
> at
> java.util.concurrent.CompletableFuture$Completion.run(CompletableFuture.java:456)
> at
>
