Distributed Publish Subscribe in Cluster

You are viewing the documentation for the new actor APIs, to view the Akka Classic documentation, see Classic Distributed Publish Subscribe.

Module info

The distributed publish subscribe topic API is available and usable with the core akka-actor-typed module, however it will only be distributed when used in a clustered application:

sbt
val AkkaVersion = "2.6.21"
libraryDependencies += "com.typesafe.akka" %% "akka-cluster-typed" % AkkaVersion
Maven
<properties>
  <scala.binary.version>2.13</scala.binary.version>
</properties>
<dependencyManagement>
  <dependencies>
    <dependency>
      <groupId>com.typesafe.akka</groupId>
      <artifactId>akka-bom_${scala.binary.version}</artifactId>
      <version>2.6.21</version>
      <type>pom</type>
      <scope>import</scope>
    </dependency>
  </dependencies>
</dependencyManagement>
<dependencies>
  <dependency>
    <groupId>com.typesafe.akka</groupId>
    <artifactId>akka-cluster-typed_${scala.binary.version}</artifactId>
  </dependency>
</dependencies>
Gradle
def versions = [
  ScalaBinary: "2.13"
]
dependencies {
  implementation platform("com.typesafe.akka:akka-bom_${versions.ScalaBinary}:2.6.21")

  implementation "com.typesafe.akka:akka-cluster-typed_${versions.ScalaBinary}"
}

The Topic Actor

Distributed publish subscribe is achieved by representing each pub sub topic with an actor, akka.actor.typed.pubsub.Topicakka.actor.typed.pubsub.Topic.

The topic actor needs to run on each node where subscribers will live or that wants to publish messages to the topic.

The identity of the topic is a tuple of the type of messages that can be published and a string topic name but it is recommended to not define multiple topics with different types and the same topic name.

Scala
sourceimport akka.actor.typed.pubsub.Topic

Behaviors.setup { context =>
  val topic = context.spawn(Topic[Message]("my-topic"), "MyTopic")
Java
sourceimport akka.actor.typed.pubsub.Topic;

Behaviors.setup(
    context -> {
      ActorRef<Topic.Command<Message>> topic =
          context.spawn(Topic.create(Message.class, "my-topic"), "MyTopic");

Local actors can then subscribe to the topic (and unsubscribe from it):

Scala
sourcetopic ! Topic.Subscribe(subscriberActor)

topic ! Topic.Unsubscribe(subscriberActor)
Java
sourcetopic.tell(Topic.subscribe(subscriberActor));

topic.tell(Topic.unsubscribe(subscriberActor));

And publish messages to the topic:

Scala
sourcetopic ! Topic.Publish(Message("Hello Subscribers!"))
Java
sourcetopic.tell(Topic.publish(new Message("Hello Subscribers!")));

Pub Sub Scalability

Each topic is represented by one Receptionist service key meaning that the number of topics will scale to thousands or tens of thousands but for higher numbers of topics will require custom solutions. It also means that a very high turnaround of unique topics will not work well and for such use cases a custom solution is advised.

The topic actor acts as a proxy and delegates to the local subscribers handling deduplication so that a published message is only sent once to a node regardless of how many subscribers there are to the topic on that node.

When a topic actor has no subscribers for a topic it will deregister itself from the receptionist meaning published messages for the topic will not be sent to it.

Delivery Guarantee

As in Message Delivery Reliability of Akka, message delivery guarantee in distributed pub sub modes is at-most-once delivery. In other words, messages can be lost over the wire. In addition to that the registry of nodes which have subscribers is eventually consistent meaning that subscribing an actor on one node will have a short delay before it is known on other nodes and published to.

If you are looking for at-least-once delivery guarantee, we recommend Alpakka Kafka.

Found an error in this documentation? The source code for this page can be found here. Please feel free to edit and contribute a pull request.