Multi JVM Testing - Version 2.4.20

Multi JVM Testing

Supports running applications (objects with main methods) and ScalaTest tests in multiple JVMs at the same time. Useful for integration testing where multiple systems communicate with each other.


The multi-JVM testing is an sbt plugin that you can find at

You can add it as a plugin by adding the following to your project/plugins.sbt:

addSbtPlugin("com.typesafe.sbt" % "sbt-multi-jvm" % "0.3.8")

You can then add multi-JVM testing to build.sbt or project/Build.scala by including the MultiJvm settings and config. Please note that MultiJvm test sources are located in src/multi-jvm/..., and not in src/test/....

Here is an example build.sbt file for sbt 0.13 that uses the MultiJvm plugin:

import com.typesafe.sbt.SbtMultiJvm
import com.typesafe.sbt.SbtMultiJvm.MultiJvmKeys.MultiJvm

val akkaVersion = "2.4.20"

val project = Project(
    id = "akka-sample-multi-node-scala",
    base = file(".")
  .settings(SbtMultiJvm.multiJvmSettings: _*)
    name := "akka-sample-multi-node-scala",
    version := "2.4.20",
    scalaVersion := "2.11.8",
    libraryDependencies ++= Seq(
      "com.typesafe.akka" %% "akka-actor" % akkaVersion,
      "com.typesafe.akka" %% "akka-remote" % akkaVersion,
      "com.typesafe.akka" %% "akka-multi-node-testkit" % akkaVersion,
      "org.scalatest" %% "scalatest" % "2.2.1" % "test"),
    // make sure that MultiJvm test are compiled by the default test compilation
    compile in MultiJvm <<= (compile in MultiJvm) triggeredBy (compile in Test),
    // disable parallel tests
    parallelExecution in Test := false,
    // make sure that MultiJvm tests are executed by the default test target,
    // and combine the results from ordinary test and multi-jvm tests
    executeTests in Test <<= (executeTests in Test, executeTests in MultiJvm) map {
      case (testResults, multiNodeResults)  =>
        val overall =
          if ( <
          testResults.summaries ++ multiNodeResults.summaries)
    licenses := Seq(("CC0", url("")))
  .configs (MultiJvm)

You can specify JVM options for the forked JVMs:

jvmOptions in MultiJvm := Seq("-Xmx256M")

Running tests

The multi-JVM tasks are similar to the normal tasks: test, test-only, and run, but are under the multi-jvm configuration.

So in Akka, to run all the multi-JVM tests in the akka-remote project use (at the sbt prompt):


Or one can change to the akka-remote-tests project first, and then run the tests:

project akka-remote-tests

To run individual tests use test-only:

multi-jvm:test-only akka.remote.RandomRoutedRemoteActor

More than one test name can be listed to run multiple specific tests. Tab-completion in sbt makes it easy to complete the test names.

It's also possible to specify JVM options with test-only by including those options after the test names and --. For example:

multi-jvm:test-only akka.remote.RandomRoutedRemoteActor -- -Dsome.option=something

Creating application tests

The tests are discovered, and combined, through a naming convention. MultiJvm test sources are located in src/multi-jvm/.... A test is named with the following pattern:


That is, each test has MultiJvm in the middle of its name. The part before it groups together tests/applications under a single TestName that will run together. The part after, the NodeName, is a distinguishing name for each forked JVM.

So to create a 3-node test called Sample, you can create three applications like the following:

package sample

object SampleMultiJvmNode1 {
  def main(args: Array[String]) {
    println("Hello from node 1")

object SampleMultiJvmNode2 {
  def main(args: Array[String]) {
    println("Hello from node 2")

object SampleMultiJvmNode3 {
  def main(args: Array[String]) {
    println("Hello from node 3")

When you call multi-jvm:run sample.Sample at the sbt prompt, three JVMs will be spawned, one for each node. It will look like this:

> multi-jvm:run sample.Sample
[info] * sample.Sample
[JVM-1] Hello from node 1
[JVM-2] Hello from node 2
[JVM-3] Hello from node 3
[success] Total time: ...

Changing Defaults

You can change the name of the multi-JVM test source directory by adding the following configuration to your project:

unmanagedSourceDirectories in MultiJvm <<=
   Seq(baseDirectory(_ / "src/some_directory_here")).join

You can change what the MultiJvm identifier is. For example, to change it to ClusterTest use the multiJvmMarker setting:

multiJvmMarker in MultiJvm := "ClusterTest"

Your tests should now be named {TestName}ClusterTest{NodeName}.

Configuration of the JVM instances

You can define specific JVM options for each of the spawned JVMs. You do that by creating a file named after the node in the test with suffix .opts and put them in the same directory as the test.

For example, to feed the JVM options -Dakka.remote.port=9991 and -Xmx256m to the SampleMultiJvmNode1 let's create three *.opts files and add the options to them. Separate multiple options with space.


-Dakka.remote.port=9991 -Xmx256m


-Dakka.remote.port=9992 -Xmx256m


-Dakka.remote.port=9993 -Xmx256m


There is also support for creating ScalaTest tests rather than applications. To do this use the same naming convention as above, but create ScalaTest suites rather than objects with main methods. You need to have ScalaTest on the classpath. Here is a similar example to the one above but using ScalaTest:

package sample

import org.scalatest.WordSpec
import org.scalatest.matchers.MustMatchers

class SpecMultiJvmNode1 extends WordSpec with MustMatchers {
  "A node" should {
    "be able to say hello" in {
      val message = "Hello from node 1"
      message must be("Hello from node 1")

class SpecMultiJvmNode2 extends WordSpec with MustMatchers {
  "A node" should {
    "be able to say hello" in {
      val message = "Hello from node 2"
      message must be("Hello from node 2")

To run just these tests you would call multi-jvm:test-only sample.Spec at the sbt prompt.

Multi Node Additions

There has also been some additions made to the SbtMultiJvm plugin to accommodate the experimental module multi node testing, described in that section.