akka : terminationCallbacks not received

Question

I have an actorsystem where i have registered some code on actorsystem termination. The code goes as follows -

object ChildActor {
  class Msg
}

class ChildActor extends Actor {
  val log = Logging(context.system, this)

  override def receive: Receive = {
    case a : String => log.info(s"received $a")
    case _ : Msg => throw new OutOfMemoryError("error")
  }
}

class ParentActor extends Actor {
  override val supervisorStrategy = OneForOneStrategy() {
    case _: OutOfMemoryError => {
      Stop
    }
    case _: Exception => Stop
  }
  val child = context.actorOf(Props[ChildActor], "child")
  override def receive: Receive = {
    case a: String => child ! a
    case msg : Msg => child ! msg
  }
}

object Test extends App {

  val customConf = ConfigFactory.parseString("""
  akka {
    jvm-exit-on-fatal-error = false
  }
  """)


  val actorSystem = ActorSystem("OOMException", ConfigFactory.load(customConf))

  actorSystem.registerOnTermination(println("going off"))

  val actor = actorSystem.actorOf(Props[ParentActor], "parentActor")

  actor ! new Msg

}

The wierd behaviour that i noticing is with my callbacks. When i run the app normally i never ever get the callback for my registered code.

However when i run in debug mode and i wait for few seconds inside the terminate() method then the line gets printed.. I am not sure why this few seconds of wait is making sure that all registered callbacks are fired whereas in the normal run mode it's not happening..Any idea anyone??

P.S : I know that OOM is fatal and but i am just curious about the problem in hand and OOM is just one scenario.

When i take a thread dump at the end i see that following two threads are always present -

"OOMException-akka.actor.default-dispatcher-2" #12 prio=5 os_prio=0 tid=0x000000001e8dd800 nid=0x1029c waiting on condition [0x0000000020eff000]
   java.lang.Thread.State: WAITING (parking)
    at sun.misc.Unsafe.park(Native Method)
    - parking to wait for  <0x0000000770700c90> (a akka.dispatch.ForkJoinExecutorConfigurator$AkkaForkJoinPool)
    at akka.dispatch.forkjoin.ForkJoinPool.scan(ForkJoinPool.java:2075)
    at akka.dispatch.forkjoin.ForkJoinPool.runWorker(ForkJoinPool.java:1979)
    at akka.dispatch.forkjoin.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:107)

"OOMException-scheduler-1" #11 prio=5 os_prio=0 tid=0x000000001e5d8800 nid=0xfba4 waiting on condition [0x000000001f35e000]
   java.lang.Thread.State: TIMED_WAITING (sleeping)
    at java.lang.Thread.sleep(Native Method)
    at akka.actor.LightArrayRevolverScheduler.waitNanos(LightArrayRevolverScheduler.scala:85)
    at akka.actor.LightArrayRevolverScheduler$$anon$4.nextTick(LightArrayRevolverScheduler.scala:265)
    at akka.actor.LightArrayRevolverScheduler$$anon$4.run(LightArrayRevolverScheduler.scala:235)
    at java.lang.Thread.run(Thread.java:748)

The scheduler should have finished as the actor system terminated ..not sure why this is in a timed waiting state..

Answer 1

App could be tricky here, maybe the JVM is terminated (because all threads left are daemon threads) before your callback can be executed.

I would first suggest putting your app code in a def main(...) method.

The use something like this instead of a println:

val promise = Promise.empty[Unit]
actorSystem.registerOnTermination(promise.success(()))
val actor = actorSystem.actorOf(Props[ParentActor], "parentActor")
actor ! new Msg
Await.ready(promise.future, Duration.INF)
println("going off")