Scheduler shutdown does not cancel scheduled actions but seems to flush them
ido
But it seems as if the next one is still executed and actually
immediately after shutdown. I think I would have to go thought all
actors or scedules and cancel them myself before calling shutdown.
Is this on purpose?
thanks,
ido
val system = ActorSystem("test")
import java.util.concurrent.TimeUnit._
system.scheduler.schedule( Duration(10, SECONDS), Duration(50,
SECONDS), new Runnable(){
override def run(){
println("still running delay")
}
}
)
system.scheduler.schedule( Duration(0, MILLISECONDS), Duration(10,
SECONDS), new Runnable(){
override def run(){
println("still running immediate")
}
}
)
Thread.sleep(100)
system.shutdown()
println("shutdown")
still running immediate
shutdown
still running delay
still running immediate
Patrik Nordwall
/Patrik
> --
> You received this message because you are subscribed to the Google Groups "Akka User List" group.
> To post to this group, send email to akka...@googlegroups.com.
> To unsubscribe from this group, send email to akka-user+...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/akka-user?hl=en.
>
rkuhn
upon shutdown:
- it must execute those vital tasks which were already queued
- it must stop accepting new tasks, otherwise there would be ample opportunity for endless loops
Hence you must cancel those tasks which shall not be run in case of system shutdown, best opportunity is the postStop() hook of an actor which has the necessary “big picture”.
Regards,
Roland
> To unsubscribe from this group, send email to akka-user+unsubscribe@googlegroups.com.
ido
Interestingly the first task which was never executed
is not executed upon shutdown.
I need to shut this down because its a database task and at some point
I have to close the connection without anything
trying to access it again.
best,
ido
def testShutdown(){
import java.util.concurrent.TimeUnit._
cancels += system.scheduler.schedule( Duration(10, SECONDS),
Duration(5, SECONDS), new Runnable(){
override def run(){
System.err.println("still running delay")
}
}
)
cancels += system.scheduler.schedule( Duration(0, MILLISECONDS),
Duration(1, SECONDS), new Runnable(){
override def run(){
System.err.println("still running immediate")
}
}
)
Thread.sleep(1000 * 2)
cancels.foreach(_.cancel)
Thread.sleep(1000 * 2)
system.shutdown()
System.err.println("shutdown")
On Mar 11, 2:26 pm, rkuhn <goo...@rkuhn.info> wrote:
> If memory serves the specific reason was that dispatchers schedule their
> shutdown, and if we lose that task we’ll leak thread pools. In general, the
> current semantics were the only one I found which does not (immediately)
> drive me into madness:
>
> upon shutdown:
> - it must execute those vital tasks which were already queued
> - it must stop accepting new tasks, otherwise there would be ample
> opportunity for endless loops
>
> Hence you must cancel those tasks which shall not be run in case of system
> shutdown, best opportunity is the postStop() hook of an actor which has the
> necessary “big picture”.
>
> Regards,
>
> Roland
>
> Am Sonntag, 11. März 2012 13:46:17 UTC+1 schrieb Patrik Nordwall:
>
>
>
>
>
>
>
>
>
> > I can confirm that all scheduled tasks are run when shutdown. I think it
> > was needed for a clean shutdown of the ActorSystem (dispatchers), but I
> > agree that it can be rather strange to run something that was scheduled
> > some (long) time in the future.
>
> > /Patrik
>
Roland Kuhn
On Mar 11, 2012, at 15:00 , ido wrote:
> Is it possible that even with cancel I get something executed?
Well, obviously there is a race between cancellation and running: if it has started running, you cannot cancel it anymore.
> Interestingly the first task which was never executed
> is not executed upon shutdown.
>
You mean in the code below? You canceled it, hence it will not be run.
> I need to shut this down because its a database task and at some point
> I have to close the connection without anything
> trying to access it again.
>
The code below prints
scala> testShutdown
still running immediate
still running immediate
shutdown
and if I remove the cancels.foreach() and the following sleep():
scala> testShutdown
still running immediate
still running immediate
shutdown
scala> still running immediate
still running delay
Here you can see that the tasks are executed exactly once after the shutdown is initiated. Keep in mind that the scheduler is shutdown only after all actors in the ActorSystem have stopped (only four in this case), not necessarily before system.shutdown() returns. You can use system.awaitTermination() if you need to be sure.
Regards,
Roland
Roland Kuhn
Typesafe – The software stack for applications that scale.
twitter: @rolandkuhn
jxtps
AkkaSystem.isTerminated doesn't seem to be it - is there anything that can give me an AkkaSystem.isTerminating?
jxtps
object AkkaShutdownMonitor {
@volatile var isTerminating = false
Akka.system.registerOnTermination {
isTerminating = true
}
}
Appears to work (i.e. gating inside the task on this isTerminating appears to solve it in dev mode in my Play 2 app).
I realize that the shutdown behavior is documented at the top of the Scheduler trait, but it would be great to see a comment about it on the actual schedule* methods - this lead to some most unfortunate issues in production.
Roland Kuhn
--
>>>>>>>>>> Read the docs: http://akka.io/docs/
>>>>>>>>>> Check the FAQ: http://doc.akka.io/docs/akka/current/additional/faq.html
>>>>>>>>>> Search the archives: https://groups.google.com/group/akka-user
---
You received this message because you are subscribed to the Google Groups "Akka User List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to akka-user+...@googlegroups.com.
Visit this group at http://groups.google.com/group/akka-user.
For more options, visit https://groups.google.com/d/optout.
Dr. Roland Kuhn
Akka Tech Lead
Typesafe – Reactive apps on the JVM.
twitter: @rolandkuhn