[
https://issues.apache.org/jira/browse/FLINK-16510?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17164358#comment-17164358
]
Till Rohrmann commented on FLINK-16510:
---------------------------------------
Thanks for providing the stack trace [~mxm]. I assume that this problem is
somewhat reproducible, right? I am asking because it would be helpful to obtain
the mixed stack trace via {{jstack -m}} which contains the JVM threads.
Moreover, in order to better understand the problem, it would also be helpful
to see the exact command with which you start the JVM process.
I agree that an easy solution would be to make Flink's exit behaviour
configurable (per default {{System.exit()}} but if users wish then to use
{{Runtime.halt()}}). However, before doing this, I would really like to better
understand why the system gets stuck in your case because {{Runtime.halt()}} is
actually being called in your case:
{code}
Thread 18575: (state = BLOCKED)
- java.lang.Shutdown.$$YJP$$halt0(int) @bci=0 (Interpreted frame)
- java.lang.Shutdown.halt0(int) @bci=14 (Interpreted frame)
- java.lang.Shutdown.halt(int) @bci=30, line=139 (Interpreted frame)
- java.lang.Runtime.halt(int) @bci=57, line=275 (Interpreted frame)
- org.apache.flink.runtime.util.JvmShutdownSafeguard$DelayedTerminator.run()
@bci=59, line=86 (Interpreted frame)
- java.lang.Thread.run() @bci=49, line=748 (Compiled frame)
{code}
The problem seems to be that all Java threads are currently being {{BLOCKED}}.
Maybe it's because a safepoint has been reached and there is some problem
coming back from it. The following JVM options could help with the debugging:
-XX:+PrintGCApplicationStoppedTime – this will actually report pause time for
all safepoints (GC related or not). Unfortunately output from this option lacks
timestamps, but it is still useful to narrow down problem to safepoints.
-XX:+PrintSafepointStatistics –XX:PrintSafepointStatisticsCount=1 – this two
options will force JVM to report reason and timings after each safepoint (it
will be reported to stdout, not GC log).
> Task manager safeguard shutdown may not be reliable
> ---------------------------------------------------
>
> Key: FLINK-16510
> URL: https://issues.apache.org/jira/browse/FLINK-16510
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Task
> Reporter: Maximilian Michels
> Assignee: Maximilian Michels
> Priority: Major
> Attachments: stack2-1.txt
>
>
> The {{JvmShutdownSafeguard}} does not always succeed but can hang when
> multiple threads attempt to shutdown the JVM. Apparently mixing
> {{System.exit()}} with ShutdownHooks and forcefully terminating the JVM via
> {{Runtime.halt()}} does not play together well:
> {noformat}
> "Jvm Terminator" #22 daemon prio=5 os_prio=0 tid=0x00007fb8e82f2800
> nid=0x5a96 runnable [0x00007fb35cffb000]
> java.lang.Thread.State: RUNNABLE
> at java.lang.Shutdown.$$YJP$$halt0(Native Method)
> at java.lang.Shutdown.halt0(Shutdown.java)
> at java.lang.Shutdown.halt(Shutdown.java:139)
> - locked <0x000000047ed67638> (a java.lang.Shutdown$Lock)
> at java.lang.Runtime.halt(Runtime.java:276)
> at
> org.apache.flink.runtime.util.JvmShutdownSafeguard$DelayedTerminator.run(JvmShutdownSafeguard.java:86)
> at java.lang.Thread.run(Thread.java:748)
> Locked ownable synchronizers:
> - None
> "FlinkCompletableFutureDelayScheduler-thread-1" #18154 daemon prio=5
> os_prio=0 tid=0x00007fb708a7d000 nid=0x5a8a waiting for monitor entry
> [0x00007fb289d49000]
> java.lang.Thread.State: BLOCKED (on object monitor)
> at java.lang.Shutdown.halt(Shutdown.java:139)
> - waiting to lock <0x000000047ed67638> (a java.lang.Shutdown$Lock)
> at java.lang.Shutdown.exit(Shutdown.java:213)
> - locked <0x000000047edb7348> (a java.lang.Class for java.lang.Shutdown)
> at java.lang.Runtime.exit(Runtime.java:110)
> at java.lang.System.exit(System.java:973)
> at
> org.apache.flink.runtime.taskexecutor.TaskManagerRunner.terminateJVM(TaskManagerRunner.java:266)
> at
> org.apache.flink.runtime.taskexecutor.TaskManagerRunner.lambda$onFatalError$1(TaskManagerRunner.java:260)
> at
> org.apache.flink.runtime.taskexecutor.TaskManagerRunner$$Lambda$27464/1464672548.accept(Unknown
> Source)
> at
> java.util.concurrent.CompletableFuture.uniWhenComplete(CompletableFuture.java:774)
> at
> java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(CompletableFuture.java:750)
> at
> java.util.concurrent.CompletableFuture.postComplete(CompletableFuture.java:488)
> at
> java.util.concurrent.CompletableFuture.completeExceptionally(CompletableFuture.java:1990)
> at
> org.apache.flink.runtime.concurrent.FutureUtils$Timeout.run(FutureUtils.java:943)
> at
> org.apache.flink.runtime.concurrent.DirectExecutorService.execute(DirectExecutorService.java:211)
> at
> org.apache.flink.runtime.concurrent.FutureUtils.lambda$orTimeout$11(FutureUtils.java:361)
> at
> org.apache.flink.runtime.concurrent.FutureUtils$$Lambda$27435/159015392.run(Unknown
> Source)
> at
> java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
> at java.util.concurrent.FutureTask.run(FutureTask.java:266)
> at
> java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$201(ScheduledThreadPoolExecutor.java:180)
> at
> java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:293)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> at java.lang.Thread.run(Thread.java:748)
> Locked ownable synchronizers:
> - <0x00000006d5e56bd0> (a
> java.util.concurrent.ThreadPoolExecutor$Worker)
> {noformat}
> Note that under this condition the JVM should terminate but it still hangs.
> Sometimes it quits after several minutes.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
