Actors

Defining an Actor class

Actor classes are implemented by extending the Actor class and implementing the receive method. The receive method should define a series of case statements (which has the type PartialFunction[Any, Unit]) that defines which messages your Actor can handle, using standard Scala pattern matching, along with the implementation of how the messages should be processed.

Here is an example:

import akka.actor.Actor
import akka.actor.Props
import akka.event.Logging

class MyActor extends Actor {
  val log = Logging(context.system, this)
  def receive = {
    case "test" => log.info("received test")
    case _      => log.info("received unknown message")
  }
}

Please note that the Akka Actor receive message loop is exhaustive, which is different compared to Erlang and the late Scala Actors. This means that you need to provide a pattern match for all messages that it can accept and if you want to be able to handle unknown messages then you need to have a default case as in the example above. Otherwise an akka.actor.UnhandledMessage(message, sender, recipient) will be published to the ActorSystem's EventStream.

Note further that the return type of the behavior defined above is Unit; if the actor shall reply to the received message then this must be done explicitly as explained below.

The result of the receive method is a partial function object, which is stored within the actor as its “initial behavior”, see Become/Unbecome for further information on changing the behavior of an actor after its construction.

Props

Props is a configuration class to specify options for the creation of actors, think of it as an immutable and thus freely shareable recipe for creating an actor including associated deployment information (e.g. which dispatcher to use, see more below). Here are some examples of how to create a Props instance.

import akka.actor.Props

val props1 = Props[MyActor]
val props2 = Props(new ActorWithArgs("arg")) // careful, see below
val props3 = Props(classOf[ActorWithArgs], "arg")

The second variant shows how to pass constructor arguments to the Actor being created, but it should only be used outside of actors as explained below.

The last line shows a possibility to pass constructor arguments regardless of the context it is being used in. The presence of a matching constructor is verified during construction of the Props object, resulting in an IllegalArgumentEception if no or multiple matching constructors are found.

// NOT RECOMMENDED within another actor:
// encourages to close over enclosing class
val props7 = Props(new MyActor)

object DemoActor {
  /**
   * Create Props for an actor of this type.
   * @param magciNumber The magic number to be passed to this actor’s constructor.
   * @return a Props for creating this actor, which can then be further configured
   *         (e.g. calling `.withDispatcher()` on it)
   */
  def props(magicNumber: Int): Props = Props(new DemoActor(magicNumber))
}

class DemoActor(magicNumber: Int) extends Actor {
  def receive = {
    case x: Int => sender() ! (x + magicNumber)
  }
}

class SomeOtherActor extends Actor {
  // Props(new DemoActor(42)) would not be safe
  context.actorOf(DemoActor.props(42), "demo")
  // ...
}

Creating Actors with Props

Actors are created by passing a Props instance into the actorOf factory method which is available on ActorSystem and ActorContext.

import akka.actor.ActorSystem

// ActorSystem is a heavy object: create only one per application
val system = ActorSystem("mySystem")
val myActor = system.actorOf(Props[MyActor], "myactor2")

Using the ActorSystem will create top-level actors, supervised by the actor system’s provided guardian actor, while using an actor’s context will create a child actor.

class FirstActor extends Actor {
  val child = context.actorOf(Props[MyActor], name = "myChild")
  // plus some behavior ...
}

It is recommended to create a hierarchy of children, grand-children and so on such that it fits the logical failure-handling structure of the application, see Actor Systems.

The call to actorOf returns an instance of ActorRef. This is a handle to the actor instance and the only way to interact with it. The ActorRef is immutable and has a one to one relationship with the Actor it represents. The ActorRef is also serializable and network-aware. This means that you can serialize it, send it over the wire and use it on a remote host and it will still be representing the same Actor on the original node, across the network.

The name parameter is optional, but you should preferably name your actors, since that is used in log messages and for identifying actors. The name must not be empty or start with $, but it may contain URL encoded characters (eg. %20 for a blank space). If the given name is already in use by another child to the same parent an InvalidActorNameException is thrown.

Actors are automatically started asynchronously when created.

Dependency Injection

If your Actor has a constructor that takes parameters then those need to be part of the Props as well, as described above. But there are cases when a factory method must be used, for example when the actual constructor arguments are determined by a dependency injection framework.

import akka.actor.IndirectActorProducer

class DependencyInjector(applicationContext: AnyRef, beanName: String)
  extends IndirectActorProducer {

  override def actorClass = classOf[Actor]
  override def produce =
    // obtain fresh Actor instance from DI framework ...
}

val actorRef = system.actorOf(
  Props(classOf[DependencyInjector], applicationContext, "hello"),
  "helloBean")

Techniques for dependency injection and integration with dependency injection frameworks are described in more depth in the Using Akka with Dependency Injection guideline and the Akka Java Spring tutorial in Typesafe Activator.

The Inbox

When writing code outside of actors which shall communicate with actors, the ask pattern can be a solution (see below), but there are two thing it cannot do: receiving multiple replies (e.g. by subscribing an ActorRef to a notification service) and watching other actors’ lifecycle. For these purposes there is the Inbox class:

implicit val i = inbox()
echo ! "hello"
i.receive() should be("hello")

There is an implicit conversion from inbox to actor reference which means that in this example the sender reference will be that of the actor hidden away within the inbox. This allows the reply to be received on the last line. Watching an actor is quite simple as well:

val target = // some actor
val i = inbox()
i watch target

Actor Lifecycle

A path in an actor system represents a "place" which might be occupied by a living actor. Initially (apart from system initialized actors) a path is empty. When actorOf() is called it assigns an incarnation of the actor described by the passed Props to the given path. An actor incarnation is identified by the path and a UID. A restart only swaps the Actor instance defined by the Props but the incarnation and hence the UID remains the same.

The lifecycle of an incarnation ends when the actor is stopped. At that point the appropriate lifecycle events are called and watching actors are notified of the termination. After the incarnation is stopped, the path can be reused again by creating an actor with actorOf(). In this case the name of the new incarnation will be the same as the previous one but the UIDs will differ.

An ActorRef always represents an incarnation (path and UID) not just a given path. Therefore if an actor is stopped and a new one with the same name is created an ActorRef of the old incarnation will not point to the new one.

ActorSelection on the other hand points to the path (or multiple paths if wildcards are used) and is completely oblivious to which incarnation is currently occupying it. ActorSelection cannot be watched for this reason. It is possible to resolve the current incarnation's ActorRef living under the path by sending an Identify message to the ActorSelection which will be replied to with an ActorIdentity containing the correct reference (see Identifying Actors via Actor Selection). This can also be done with the resolveOne method of the ActorSelection, which returns a Future of the matching ActorRef.

Lifecycle Monitoring aka DeathWatch

In order to be notified when another actor terminates (i.e. stops permanently, not temporary failure and restart), an actor may register itself for reception of the Terminated message dispatched by the other actor upon termination (see Stopping Actors). This service is provided by the DeathWatch component of the actor system.

Registering a monitor is easy:

import akka.actor.{ Actor, Props, Terminated }

class WatchActor extends Actor {
  val child = context.actorOf(Props.empty, "child")
  context.watch(child) // <-- this is the only call needed for registration
  var lastSender = system.deadLetters

  def receive = {
    case "kill" =>
      context.stop(child); lastSender = sender()
    case Terminated(`child`) => lastSender ! "finished"
  }
}

It should be noted that the Terminated message is generated independent of the order in which registration and termination occur. In particular, the watching actor will receive a Terminated message even if the watched actor has already been terminated at the time of registration.

Registering multiple times does not necessarily lead to multiple messages being generated, but there is no guarantee that only exactly one such message is received: if termination of the watched actor has generated and queued the message, and another registration is done before this message has been processed, then a second message will be queued, because registering for monitoring of an already terminated actor leads to the immediate generation of the Terminated message.

It is also possible to deregister from watching another actor’s liveliness using context.unwatch(target). This works even if the Terminated message has already been enqueued in the mailbox; after calling unwatch no Terminated message for that actor will be processed anymore.

Start Hook

Right after starting the actor, its preStart method is invoked.

override def preStart() {
  child = context.actorOf(Props[MyActor], "child")
}

This method is called when the actor is first created. During restarts it is called by the default implementation of postRestart, which means that by overriding that method you can choose whether the initialization code in this method is called only exactly once for this actor or for every restart. Initialization code which is part of the actor’s constructor will always be called when an instance of the actor class is created, which happens at every restart.

Restart Hooks

All actors are supervised, i.e. linked to another actor with a fault handling strategy. Actors may be restarted in case an exception is thrown while processing a message (see Supervision and Monitoring). This restart involves the hooks mentioned above:

1. The old actor is informed by calling preRestart with the exception which caused the restart and the message which triggered that exception; the latter may be None if the restart was not caused by processing a message, e.g. when a supervisor does not trap the exception and is restarted in turn by its supervisor, or if an actor is restarted due to a sibling’s failure. If the message is available, then that message’s sender is also accessible in the usual way (i.e. by calling sender).

   This method is the best place for cleaning up, preparing hand-over to the fresh actor instance, etc. By default it stops all children and calls postStop.

2. The initial factory from the actorOf call is used to produce the fresh instance.

3. The new actor’s postRestart method is invoked with the exception which caused the restart. By default the preStart is called, just as in the normal start-up case.

An actor restart replaces only the actual actor object; the contents of the mailbox is unaffected by the restart, so processing of messages will resume after the postRestart hook returns. The message that triggered the exception will not be received again. Any message sent to an actor while it is being restarted will be queued to its mailbox as usual.

Stop Hook

After stopping an actor, its postStop hook is called, which may be used e.g. for deregistering this actor from other services. This hook is guaranteed to run after message queuing has been disabled for this actor, i.e. messages sent to a stopped actor will be redirected to the deadLetters of the ActorSystem.

Tell: Fire-forget

This is the preferred way of sending messages. No blocking waiting for a message. This gives the best concurrency and scalability characteristics.

actorRef ! message

If invoked from within an Actor, then the sending actor reference will be implicitly passed along with the message and available to the receiving Actor in its sender(): ActorRef member method. The target actor can use this to reply to the original sender, by using sender() ! replyMsg.

If invoked from an instance that is not an Actor the sender will be deadLetters actor reference by default.

Ask: Send-And-Receive-Future

The ask pattern involves actors as well as futures, hence it is offered as a use pattern rather than a method on ActorRef:

import akka.pattern.{ ask, pipe }
import system.dispatcher // The ExecutionContext that will be used
case class Result(x: Int, s: String, d: Double)
case object Request

implicit val timeout = Timeout(5 seconds) // needed for `?` below

val f: Future[Result] =
  for {
    x <- ask(actorA, Request).mapTo[Int] // call pattern directly
    s <- (actorB ask Request).mapTo[String] // call by implicit conversion
    d <- (actorC ? Request).mapTo[Double] // call by symbolic name
  } yield Result(x, s, d)

f pipeTo actorD // .. or ..
pipe(f) to actorD

This example demonstrates ask together with the pipeTo pattern on futures, because this is likely to be a common combination. Please note that all of the above is completely non-blocking and asynchronous: ask produces a Future, three of which are composed into a new future using the for-comprehension and then pipeTo installs an onComplete-handler on the future to affect the submission of the aggregated Result to another actor.

Using ask will send a message to the receiving Actor as with tell, and the receiving actor must reply with sender() ! reply in order to complete the returned Future with a value. The ask operation involves creating an internal actor for handling this reply, which needs to have a timeout after which it is destroyed in order not to leak resources; see more below.

try {
  val result = operation()
  sender() ! result
} catch {
  case e: Exception =>
    sender() ! akka.actor.Status.Failure(e)
    throw e
}

If the actor does not complete the future, it will expire after the timeout period, completing it with an AskTimeoutException. The timeout is taken from one of the following locations in order of precedence:

1. explicitly given timeout as in:

import scala.concurrent.duration._
import akka.pattern.ask
val future = myActor.ask("hello")(5 seconds)

2. implicit argument of type akka.util.Timeout, e.g.

import scala.concurrent.duration._
import akka.util.Timeout
import akka.pattern.ask
implicit val timeout = Timeout(5 seconds)
val future = myActor ? "hello"

See Futures for more information on how to await or query a future.

The onComplete, onSuccess, or onFailure methods of the Future can be used to register a callback to get a notification when the Future completes. Gives you a way to avoid blocking.

Forward message

You can forward a message from one actor to another. This means that the original sender address/reference is maintained even though the message is going through a 'mediator'. This can be useful when writing actors that work as routers, load-balancers, replicators etc.

target forward message

PoisonPill

You can also send an actor the akka.actor.PoisonPill message, which will stop the actor when the message is processed. PoisonPill is enqueued as ordinary messages and will be handled after messages that were already queued in the mailbox.

Graceful Stop

gracefulStop is useful if you need to wait for termination or compose ordered termination of several actors:

import akka.pattern.gracefulStop
import scala.concurrent.Await

try {
  val stopped: Future[Boolean] = gracefulStop(actorRef, 5 seconds, Manager.Shutdown)
  Await.result(stopped, 6 seconds)
  // the actor has been stopped
} catch {
  // the actor wasn't stopped within 5 seconds
  case e: akka.pattern.AskTimeoutException =>
}

object Manager {
  case object Shutdown
}

class Manager extends Actor {
  import Manager._
  val worker = context.watch(context.actorOf(Props[Cruncher], "worker"))

  def receive = {
    case "job" => worker ! "crunch"
    case Shutdown =>
      worker ! PoisonPill
      context become shuttingDown
  }

  def shuttingDown: Receive = {
    case "job" => sender() ! "service unavailable, shutting down"
    case Terminated(`worker`) =>
      context stop self
  }
}

When gracefulStop() returns successfully, the actor’s postStop() hook will have been executed: there exists a happens-before edge between the end of postStop() and the return of gracefulStop().

In the above example a custom Manager.Shutdown message is sent to the target actor to initiate the process of stopping the actor. You can use PoisonPill for this, but then you have limited possibilities to perform interactions with other actors before stopping the target actor. Simple cleanup tasks can be handled in postStop.

Upgrade

Akka supports hotswapping the Actor’s message loop (e.g. its implementation) at runtime: invoke the context.become method from within the Actor. become takes a PartialFunction[Any, Unit] that implements the new message handler. The hotswapped code is kept in a Stack which can be pushed and popped.

To hotswap the Actor behavior using become:

class HotSwapActor extends Actor {
  import context._
  def angry: Receive = {
    case "foo" => sender() ! "I am already angry?"
    case "bar" => become(happy)
  }

  def happy: Receive = {
    case "bar" => sender() ! "I am already happy :-)"
    case "foo" => become(angry)
  }

  def receive = {
    case "foo" => become(angry)
    case "bar" => become(happy)
  }
}

This variant of the become method is useful for many different things, such as to implement a Finite State Machine (FSM, for an example see Dining Hakkers). It will replace the current behavior (i.e. the top of the behavior stack), which means that you do not use unbecome, instead always the next behavior is explicitly installed.

The other way of using become does not replace but add to the top of the behavior stack. In this case care must be taken to ensure that the number of “pop” operations (i.e. unbecome) matches the number of “push” ones in the long run, otherwise this amounts to a memory leak (which is why this behavior is not the default).

case object Swap
class Swapper extends Actor {
  import context._
  val log = Logging(system, this)

  def receive = {
    case Swap =>
      log.info("Hi")
      become({
        case Swap =>
          log.info("Ho")
          unbecome() // resets the latest 'become' (just for fun)
      }, discardOld = false) // push on top instead of replace
  }
}

object SwapperApp extends App {
  val system = ActorSystem("SwapperSystem")
  val swap = system.actorOf(Props[Swapper], name = "swapper")
  swap ! Swap // logs Hi
  swap ! Swap // logs Ho
  swap ! Swap // logs Hi
  swap ! Swap // logs Ho
  swap ! Swap // logs Hi
  swap ! Swap // logs Ho
}

Encoding Scala Actors nested receives without accidentally leaking memory

See this Unnested receive example.

What happens to the Message

If an exception is thrown while a message is being processed (i.e. taken out of its mailbox and handed over to the current behavior), then this message will be lost. It is important to understand that it is not put back on the mailbox. So if you want to retry processing of a message, you need to deal with it yourself by catching the exception and retry your flow. Make sure that you put a bound on the number of retries since you don't want a system to livelock (so consuming a lot of cpu cycles without making progress). Another possibility would be to have a look at the PeekMailbox pattern.

What happens to the mailbox

If an exception is thrown while a message is being processed, nothing happens to the mailbox. If the actor is restarted, the same mailbox will be there. So all messages on that mailbox will be there as well.

What happens to the actor

If code within an actor throws an exception, that actor is suspended and the supervision process is started (see Supervision and Monitoring). Depending on the supervisor’s decision the actor is resumed (as if nothing happened), restarted (wiping out its internal state and starting from scratch) or terminated.

Initialization via constructor

Using the constructor for initialization has various benefits. First of all, it makes it possible to use val fields to store any state that does not change during the life of the actor instance, making the implementation of the actor more robust. The constructor is invoked for every incarnation of the actor, therefore the internals of the actor can always assume that proper initialization happened. This is also the drawback of this approach, as there are cases when one would like to avoid reinitializing internals on restart. For example, it is often useful to preserve child actors across restarts. The following section provides a pattern for this case.

Initialization via preStart

The method preStart() of an actor is only called once directly during the initialization of the first instance, that is, at creation of its ActorRef. In the case of restarts, preStart() is called from postRestart(), therefore if not overridden, preStart() is called on every incarnation. However, overriding postRestart() one can disable this behavior, and ensure that there is only one call to preStart().

One useful usage of this pattern is to disable creation of new ActorRefs for children during restarts. This can be achieved by overriding preRestart():

override def preStart(): Unit = {
  // Initialize children here
}

// Overriding postRestart to disable the call to preStart()
// after restarts
override def postRestart(reason: Throwable): Unit = ()

// The default implementation of preRestart() stops all the children
// of the actor. To opt-out from stopping the children, we
// have to override preRestart()
override def preRestart(reason: Throwable, message: Option[Any]): Unit = {
  // Keep the call to postStop(), but no stopping of children
  postStop()
}

Please note, that the child actors are still restarted, but no new ActorRef is created. One can recursively apply the same principles for the children, ensuring that their preStart() method is called only at the creation of their refs.

For more information see What Restarting Means.

Initialization via message passing

There are cases when it is impossible to pass all the information needed for actor initialization in the constructor, for example in the presence of circular dependencies. In this case the actor should listen for an initialization message, and use become() or a finite state-machine state transition to encode the initialized and uninitialized states of the actor.

var initializeMe: Option[String] = None

override def receive = {
  case "init" =>
    initializeMe = Some("Up and running")
    context.become(initialized, discardOld = true)

}

def initialized: Receive = {
  case "U OK?" => initializeMe foreach { sender() ! _ }
}

If the actor may receive messages before it has been initialized, a useful tool can be the Stash to save messages until the initialization finishes, and replaying them after the actor became initialized.