Develop Apache Storm topologies using Python on HDInsight

Learn how to create an Apache Storm topology that uses Python components. Apache Storm supports multiple languages, even allowing you to combine components from several languages in one topology. The Flux framework (introduced with Storm 0.10.0) allows you to easily create solutions that use Python components.


The information in this document was tested using Storm on HDInsight 3.5. Linux is the only operating system used on HDInsight version 3.4 or greater. For more information, see HDInsight retirement on Windows.

The code for this project is available at


  • Python 2.7 or higher

  • Java JDK 1.8 or higher

  • Maven 3

  • (Optional) A local Storm development environment. A local Storm environment is only needed if you want to run the topology locally. For more information, see Setting up a development environment.

Storm multi-language support

Apache Storm was designed to work with components written using any programming language. The components must understand how to work with the Thrift definition for Storm. For Python, a module is provided as part of the Apache Storm project that allows you to easily interface with Storm. You can find this module at

Storm is a Java process that runs on the Java Virtual Machine (JVM). Components written in other languages are executed as subprocesses. The Storm communicates with these subprocesses using JSON messages sent over stdin/stdout. More details on communication between components can be found in the Multi-lang Protocol documentation.

Python with the Flux framework

The Flux framework allows you to define Storm topologies separately from the components. The Flux framework uses YAML to define the Storm topology. The following text is an example of how to reference a Python component in the YAML document:

# Spout definitions
  - id: "sentence-spout"
    className: "org.apache.storm.flux.wrappers.spouts.FluxShellSpout"
      # Command line
      - ["python", ""]
      # Output field(s)
      - ["sentence"]
    # parallelism hint
    parallelism: 1

The class FluxShellSpout is used to start the script that implements the spout.

Flux expects the Python scripts to be in the /resources directory inside the jar file that contains the topology. So this example stores the Python scripts in the /multilang/resources directory. The pom.xml includes this file using the following XML:

<!-- include the Python components -->

As mentioned earlier, there is a file that implements the Thrift definition for Storm. The Flux framework includes automatically when the project is built, so you don't have to worry about including it.

Build the project

From the root of the project, use the following command:

mvn clean compile package

This command creates a target/WordCount-1.0-SNAPSHOT.jar file that contains the compiled topology.

Run the topology locally

To run the topology locally, use the following command:

storm jar WordCount-1.0-SNAPSHOT.jar org.apache.storm.flux.Flux -l -R /topology.yaml

This command requires a local Storm development environment. For more information, see Setting up a development environment

Once the topology starts, it emits information to the local console similar to the following text:

24302 [Thread-25-sentence-spout-executor[4 4]] INFO  o.a.s.s.ShellSpout - ShellLog pid:2436, name:sentence-spout Emiting the cow jumped over the moon
24302 [Thread-30] INFO  o.a.s.t.ShellBolt - ShellLog pid:2438, name:splitter-bolt Emitting the
24302 [Thread-28] INFO  o.a.s.t.ShellBolt - ShellLog pid:2437, name:counter-bolt Emitting years:160
24302 [Thread-17-log-executor[3 3]] INFO  o.a.s.f.w.b.LogInfoBolt - {word=the, count=599}
24303 [Thread-17-log-executor[3 3]] INFO  o.a.s.f.w.b.LogInfoBolt - {word=seven, count=302}
24303 [Thread-17-log-executor[3 3]] INFO  o.a.s.f.w.b.LogInfoBolt - {word=dwarfs, count=143}
24303 [Thread-25-sentence-spout-executor[4 4]] INFO  o.a.s.s.ShellSpout - ShellLog pid:2436, name:sentence-spout Emiting the cow jumped over the moon
24303 [Thread-30] INFO  o.a.s.t.ShellBolt - ShellLog pid:2438, name:splitter-bolt Emitting cow
24303 [Thread-17-log-executor[3 3]] INFO  o.a.s.f.w.b.LogInfoBolt - {word=four, count=160}

To stop the topology, use Ctrl + C.

Run the Storm topology on HDInsight

  1. Use the following command to copy the WordCount-1.0-SNAPSHOT.jar file to your Storm on HDInsight cluster:

    scp target\WordCount-1.0-SNAPSHOT.jar

    Replace sshuser with the SSH user for your cluster. Replace mycluster with the cluster name. You may be prompted to enter the password for the SSH user.

    For more information on using SSH and SCP, see Use SSH with HDInsight.

  2. Once the file has been uploaded, connect to the cluster using SSH:

  3. From the SSH session, use the following command to start the topology on the cluster:

    storm jar WordCount-1.0-SNAPSHOT.jar org.apache.storm.flux.Flux -r -R /topology.yaml
  4. You can use the Storm UI to view the topology on the cluster. The Storm UI is located at Replace mycluster with your cluster name.


Once started, a Storm topology runs until stopped. To stop the topology, use one of the following methods:

  • The storm kill TOPOLOGYNAME command from the command line
  • The Kill button in the Storm UI.

Next steps

See the following documents for other ways to use Python with HDInsight: