Integrating Play for Java and Akka

Integrating Play for Java and Akka

The Play Framework gives a web-driven approach to manage regulate building web applications, regardless it likewise gives an instrument to you to make on a very basic level adaptable applications. By organizing Play with Akka you can offload all CPU-bona fide dealing with another system running on another specialist. You can figure out Play with Akka to really see extraordinarily adaptable web application development.

The previous two articles in this game-plan, Introduction to Play 2 for Java, and Developing Scalable Web Applications with Play, examined the evaluation of the Play Framework, set up an improvement circumstance, made a Hello, World application, and accordingly investigated Play’s assistance for a zone driven blueprint and its use of Scala positions as we conveyed an essential Widget the heads application. In the end we direct our fixation toward clearly the most bracing bit of Play: abnormal overseeing. Here we see Play’s assistance for sending messages to “performers,” surrendering the referencing masterminding string while those performers run, and a short period of time later sparing and reestablishing a response when those performers complete. Besides, we research joining Play with Akka so our play applications can send messages to performers running in an other Akka laborer for getting ready. To spread it out, we will figure out how to help plainly more coordinated referencing than we have strings and scale our application incredibly.

Offbeat Processing

The Play Framework not simply allows you to consider seeing HTTP instead of Java APIs, which alone would be adequate to ask you to get it at any rate it nearly allows your application to give up its business masterminding string while simultaneously executing long-running exercises. For example, in a standard web structure, if you need to make twelve data base choices to satisfy you’re referencing, by then you would hinder your string while simultaneously tolerating that the data base will respond. In case your web compartment had 50 strings, by then everything considered you could fortify 50 simultaneous referencing. Play, regardless, licenses you to convey a message, send it to a “performer,” and some time later calm accommodation its string. The performer would then have the choice to make your data base choices for you and when it is finished managing, it can send your Play application a response message. The play gives the message to your application, close to the business/response portraying with the objective that you can reach back to the visitor. This derives 50 strings can fortify most likely more than 50 simultaneous arrangements. Besides, the performers to which you send messages don’t all things considered ought to be formed with your application; they can be running in an Akka laborer on another machine. This part encourages the best way to deal with execute performers in the Play JVM, and the going with zone advises the most ideal approach to execute performers on an outside laborer.

Integrating Play for Java and Akka

As of recently our controller rehearses have restored an outcome, in any case, starting at now we will change them to reestablish the certification of an outcome: Promise. Essentially this initiates we will “at long last” return a response. Play understands that when it sees an affirmation for a result that it can suspend the orchestrating of that request and reuse the string for various errands. Right when the result appears, by then Play can use a string to take out the response, convert it to a Result, and return that outcome back to the visitor.

A full depiction of Akka performers is past the degree of this article, at any rate I need to give you enough to be dangerous (and you can investigate an article that I clarified Akka here: akka.html). Akka executes the Actor Model (, which was portrayed in 1973 to support concurrent structures. Energy for the Actor Model has returned starting late with the presence of spread taking care of: In 1973 it was attempting to hover planning over different physical machines yet now we’re attempting to ignore on managing distinctive virtual machines.

Akka works through a level of indirection: All performers live in an “ActorSystem” and your application requests a reference to a performer (ActorRef) from the ActorSystem. Your application builds up a message and sends it to the ActorRef. The ActorRef gives the message to a MessageDispatcherthat thusly gives the message to the Actor’s MessageQueue. Decisively when the performer is alloted CPU time, the entertainer’s Mailbox checks its MessageQueue and, if there are messages open, the Mailbox eliminates the message from the MessageQueue and passes it to the Actor’s onReceive() technique. This is summarized in Figure 1.

Integrating Play for Java and AkkaFigure 1 Akka’s implementation of the Actor Model

Incorporating Play for Java and AkkaFigure 1 Akka’s execution of the Actor Model

The supported circumstance to this method is that we can have unlimited messages experiencing a JVM and the application won’t crash: Under excellent weight, the MessageQueue may back up, yet the performers will deal with messages using the JVM’s strings as they are fit. In addition, the indirection allows the performer’s area to be decoupled from the client. (The performer may be in the indistinct JVM or the country over running in another authority ranch.)

In this section, we have to cause a performer in the region To jvm and send it a message. Posting 1 shows the source code for our HelloLocalActor class.

Posting 1.

  • Local Hello, World Actor


open class HelloLocalActor augments UntypedActor

{@Overridepublic void onReceive( Object message ) tosses Exception{if( message instanceof MyMessage ){MyMessage myMessage = ( MyMessage )message;myMessage.setMessage( “Neighborhood Hello, ” + myMessage.getMessage() );getSender().tell( myMessage, getSelf() );}else{unhandled( message );}}


Entertainers augment UntypedActor and deny its onReceive() technique, which is passed an Object. It normally studies such a message and thus either handle the message or returns unhandled(message). The message that we’re going around is of type MyMessage, which wraps a singular String property named message (showed up in Listing 2.) If the message is of type my message, by then the HelloLocalActor prefixes the message with “Close to Hello, ” and illuminates our sender by conjuring getSender().tell().getSender() restores a reference to the Actor that sent the message and tell() is the instrument through which we can send a response message. The tell() technique sees the message to send correspondingly as a wellspring of point of view to the sender of the message, which for this circumstance is the HelloLocalActor.

Listing 2.

pack com.geekcap.informit.akka;import;public class MyMessage executes Serializable

{private String message;public MyMessage(){}public MyMessage( String message ){this.message = message;}public String getMessage(){return message;}public void setMessage( String message ){this.message = message;}


Since we have a performer that can strategy a MyMessage, we ought to merge a controller action that can call it. Posting 3 shows the source code for the essential kind of our Application class, which contains a localHello() activity.

Listing 3.

bunch controllers;import;



import play.*;

import play.libs.Akka;

import play.libs.F.Promise;

import play.libs.F.Function;

import play.mvc.*;

import views.html.*;

import static akka.pattern.Patterns.ask;import actors.HelloLocalActor;

import com.geekcap.informit.akka.MyMessage;public class Application creates Controller {static actorSystem = ActorSystem.create( “play” );static {//Create our neighborhood actorsactorSystem.actorOf( Props.create( HelloLocalActor.class ), “HelloLocalActor” );}public static Result record() {return ok(index.render(“Your new application is ready.”));}/* Controller movement that builds up a MyMessage and sends it to our local* Hello, World entertainer @param nameThe name of the person to greet* @returnThe insistence of a Result*/open static Promise localHello( String name ){//Look up the actorActorSelection myActor =actorSystem.actorSelection( “customer/HelloLocalActor” );//Connstruct our messageMyMessage message = new MyMessage( name );//As the entertainer for a response to the message (and a 30 second break);//ask reestablishes an Akka Future, so we wrap it with a Play Promisereturn Promise.wrap(ask(myActor, message, 30000)).map(new Function() {public Result apply(Object response) {if( response instanceof MyMessage ) {MyMessage message = ( MyMessage )response;return OK( message.getMessage() );}return notFound( “Message isn’t of type MyMessage” );}});}


The Application class contains a static reference to the ActorSystem and presents it as its portrayed. We need the ActorSystem to have our entertainers correspondingly as to send and get messages. The ActorSystemis named, which makes it addressable, and it makes it possible to have different ActorSystems in the equal JVM. For our condition we named our ActorSystem “play”, yet you may have fundamentally named it “foo” or “bar”. Besides, there is a static code impede in which we make the HelloLocalActor. We make entertainers by conjuring the entertainer() methodology on the ActorSystem (there are various frameworks, at any rate this is obviously one of the most straightforward), passing it a Props object with the class that executes the entertainer. We besides pass the entertainer() framework the name of the entertainer with the objective that it will be more straightforward for us to look at later.

Right when the localHello() development is brought, we search for our entertainer, by name, using the ActorSystem’s entertainer choice() technique. Entertainers are seen using an entertainer way, which is of the alliance:


For this condition, we are sifting for an entertainer in the close by JVM and our ActorSystem is starting at now named, so we don’t need to show the ActorSystemName, the specialist, or the port. There are two watchmen in Akka: structure and client. The structure contains the total of Akka’s entertainers and customer contains our own. The HelloLocalActor is portrayed truly in the ActorSystem, so it is seen as a “raised level entertainer”. In the event that it by one way or another figured out some approach to make sub-entertainers of its own, by then they would be portrayed as sub-entertainers of the HelloLocalActor. Thusly, we can find our entertainer with the way “customer/HelloLocalActor”. In the going with piece, we’ll be investigating an entertainer that isn’t in our neighborhood JVM, so we’ll see a full entertainer way.

The ActorSelection is an ActorRef, so now, we basically need to manufacture a message and send it to the ActorRef. We make an essential MyMessage and by then enter the upsetting looking code. There’s an awesome approach going on in the going with line, so we ought to ponder what it’s doing a bit at a time:

Patterns.ask: This is an Akka work that grants something express non simultaneously to an entertainer (an ActorRef) with a message and a break, which as time goes on restores a reaction through

a scala.concurrent.Future

Next, we need to add another course to the courses record to send a referencing to the localHello() framework:

GET/close to howdy/: name controllers.Application.localHello( name : String )

Finally, we need to add Akka sponsorship to our headway report (build.sbt). Posting 4 shows the substance of our build.sbt record.

At last, we have to add Akka sponsorship to our advancement report (build.sbt). Posting 4 shows the substance of our build.sbt record.

Posting 4. build.sbt

name := “SimplePlayApp”version := “1.0-SNAPSHOT”libraryDependencies ++= Seq(javaJdbc,javaEbean,cache,”com.typesafe.akka” % “akka-remote_2.10” % “2.2.3”,”com.geekcap.informit.akka” % “akka-messages” % “1.0-SNAPSHOT”


We could import Akka’s performer’s social event, yet since in the going with zone we will call an external Akka authority, I picked to use Akka-far away. Note that the structure isn’t the latest: You need to sort out your Play and Akka structures. (I found the most annoying way imaginable using the latest structure and seeing unusual staggers that didn’t quick me to the way that I don’t have the correct structure.) The documentation is fairly not actually identical to a Maven POM record, at any rate the information is undefined:

pack ID % antiquated extraordinariness ID % version

You’ll see that I have a substitute undertaking for Akka-messages. We will serialize MyMessageinstances and sending them over the framework to the Akka genius (called a cut back scale part) so unfortunately the messages are unclear. Rather than reordering the code, I decided to make another undertaking that wires basically our message(s) and import that experience into both of our errands (Play and Akka).

With this total, start Play (execute Play from the sales line and accumulate the run interest from the Play brief) and open a program to http://localhost:9000/neighborhood welcome/YourName, and you should see “Hello, YourName”.

Mix with Akka

Posting 6.

pack com.geekcap.informit.akka;import;


import akka.kernel.Bootable; open class MyKernel recognizes Bootable

{final ActorSystem structure = ActorSystem.create(“mykernel”);public void shutdown(){//Shutdown our performer systemsystem.shutdown();}public void startup(){//Create our actorssystem.actorOf( Props.create( HelloWorldActor.class ), “HelloWorldActor” );}


Posting 6 makes a class considered MyKernel that comprehends the akka.kernel.Bootable interface. This interface depicts two strategies: startup() and conclusion(), which are viewed as when the segment starts up and shuts down, as it were. We make an ActorSystem named “package” as our bootable class is made and we shut it down when the conclusion() technique is called. You are permitted to name your ActorSystem anything that you need: When Play sets up a relationship with our ActorSystem, it will send the name as a cutoff in the performer way. In the startup() method we make all our colossal level performers, with their names.

To make our performer open by suggestion, we need to add an application.conf report to the base of our resultant JAR record. In Maven encounters, we can place this record in src/head/resources. Posting 7 shows the substance of the application.conf record.

Posting 6.

pack com.geekcap.informit.akka;import;


import akka.kernel.Bootable; open class MyKernel acknowledges Bootable

{final ActorSystem structure = ActorSystem.create(“mykernel”);public void shutdown(){//Shutdown our entertainer systemsystem.shutdown();}public void startup(){//Create our actorssystem.actorOf( Props.create( HelloWorldActor.class ), “HelloWorldActor” );}


Posting 6 makes a class considered MyKernel that understands the akka.kernel.Bootable interface. This interface portrays two procedures: startup() and closure(), which are seen as when the section fires up and closes down, only. We make an ActorSystem named “parcel” as our bootable class is made and we shut it down when the closure() method is called. You are allowed to name your ActorSystem anything that you need: When Play sets up an association with our ActorSystem, it will send the name as a cutoff in the entertainer way. In the startup() technique we make all our huge level entertainers, with their names.

To make our performer open by suggestion, we need to add an application.conf report to the base of our resultant JAR record. In Maven encounters, we can place this record in src/head/resources. Posting 7 shows the substance of the application.conf record.

Posting 7. application.conf

akka {actor {provider = “akka.remote.RemoteActorRefProvider”}remote {enabled-transports = [“akka.remote.netty.tcp”]netty.tcp {hostname = “”port = 2552}}


The application.conf file sets up a removed supplier that tunes in on the close to machine on port 2552, which is Akka’s default port. This structure licenses outside Akka clients to send messages to performers running in our Akka scaled a downscale piece.

Posting 8. pom.xml record for Akka Actors


Posting 8 structures our Akka entertainer and touch and bundles the application.conf archive with it into a solitary JAR record. The POM record combines the Akka entertainer and Kernel conditions yet additionally merges a reference to our Akka-messages adventure. The source code related with this article will have that experience – you’ll have to make it before you can create this task. Moreover, survey that we externalized our message to its own undertaking so we could recall it for both the Akka experience comparably as the Play experience.

You can make the endeavor with the going with request:

mvn clean present

Since you have your entertainer and part in a JAR report, you need to set up an Akka zone. You can download Akka from here ( I downloaded the previous assortment (2.2.4) to ensure that it is faultless with the interpretation of Play we presented (2.2.3). The specific assortments don’t have such an effect, basically guarantee that when you present both Play and Akka that the structures compose. Download the ZIP record and decompress it to your hard drive. Next, set the AKKA_HOME condition variable to the stock to which you decompressed the Akka annal.

Integrating Play for Java and Akka

To send your entertainer and part to Akka, duplicate the Akka-actors.jar record that you just endeavored to Akka’s pass on document and duplicate the Akka-messages.jar file (that contains the MyMessage class) to Akka’s lib/Akka vault. With these two records set up, you can dispatch Akka from the compartment vault by executing the going with demand:

./akka com.geekcap.informit.akka.kernel

After the Akka header is appeared, you should see something like the going with:

Starting up com.geekcap.informit.akka.kernel

Satisfactorily began Akka

Eventually we have to retrofit the Play application to make the far off choice to Akka. We as of late combined the Akka remoting reliance in our build.sbt record, yet to get hit we’re up moving to need to add the going with to the furthest reaches of our conf/application.conf chronicle:

akka.default-dispatcher.fork-join-executor.pool-size-max = 64 = onakka {actor {provider = “akka.remote.RemoteActorRefProvider”}remote {enabled-transports = [“akka.remote.netty.tcp”]netty.tcp {hostname = “”port = 2555}}


This will arrange Play to tune in for callbacks from Akka on port 2555. (The port number doesn’t have any sort of impact; it fundamentally should be not actually equal to the Akka port in the function that you’re running them on a relative machine.) Next, we will fuse another course and another regulator activity to our Application class. The going with shows the new course (added to the conf/courses record):

GET/howdy/:name controllers.Application.hello( name : String )

This helpers a GET mentioning to hi/:name to the welcome() development in the Application class, which is appeared in Listing 9.

Listing 9. Application Class’s hello() Method

open static Promise greetings( String name ){ActorSelection myActor =actorSystem.actorSelection(“akka.tcp://mykernel@” );MyMessage message = new MyMessage( name );return Promise.wrap(ask(myActor, message, 30000)).map(new Function() {public Result apply(Object response) {if( response instanceof MyMessage ) {MyMessage message = ( MyMessage )response;return okay( message.getMessage() );}return notFound( “Message isn’t of type MyMessage” );}});}

The welcome() procedure in Listing 9 looks basically indistinct from our localHello() methodology in Listing 3. The principle differentiation is that we changed the performer route from “customer/HelloLocalActor” to feature the HelloActor we have running in Akka:


This entertainer way can be depicted as follows:

Akka: Identifies this as an entertainer way.

TCP: Defines the call as utilizing TCP (Transmission Control Protocol), which will be made plans to Netty from the application.conf record.

partition: The name of the entertainer framework, which we portrayed in the MyKernel class in the Akka experience. The region and port of Akka.

client: The client gatekeeper, which is the guard that deals with all our raised level entertainers.

HelloWorldActor: The name of the raised level entertainer to which to send the message.

Similarly, that is it. Spare your file, start Play on the off chance that it isn’t beginning at now running, and some time later open a web program to http://localhost:9000/welcome/YourName

As a reaction, you should see “Howdy, YourName”. In the Play comfort, you should see something like the going with:

[INFO] [05/23/2014 14:34:32.395] [] [Remoting] Starting remoting [INFO] [05/23/2014 14:34:33.490] [] [Remoting] Remoting began; tuning in on addresses :[akka.tcp://play@]

This says Play has begun remoting and is tuning in for a reaction in the entertainer structure “play”, which is depicted in Listing 3, on the port the near to machine ( on port 2555, the two of which are depicted in application.conf.

In the Akka maintain you should see something like the going with:

Gotten message: com.geekcap.informit.akka.MyMessage@5a5a7c64

This is from the System.out.println() call that we made in the HelloWorldActor class.


The Play Framework not just gives a brand name web-driven point of view for making web applications, in any case, it in addition can be utilized to non at the same time measure demands while not cornering strings that are never serving to beside hanging on for long-running activities. We investigated Play’s nonconcurrent preparing model by first allocating demand managing to a near to entertainer running in the tantamount JVM and thusly building up an association with an Akka little extension piece for dealing with on a possibly unique worker. This is the detect the guaranteed intensity of Play and Akka begins from: Your Play application can get demands, dispatch work to a ton of Akka more unobtrusive extension pieces, and some time later, when that dealing with is done, it can make a reaction to deliver off the guest. Likewise, recalling that it is keeping it together for a reaction from the far away entertainers, Play can surrender the mentioning dealing with string to permit that string to support additional solicitations. Thusly, this proposes in the event that you have 50 strings in your string pool, you can fulfill evidently in excess of 50 concurrent deals!

Leave a Reply

Your email address will not be published. Required fields are marked *