Developer documentation

This document summarizes information relevant to Storm committers and contributors. It includes information about the development processes and policies as well as the tools we use to facilitate those.

Table of Contents

  • Welcome!
  • Workflows
    • Report a bug
    • Request a new feature
    • Contribute code
    • Contribute documentation
    • Pull requests
      • Create a pull request
      • Approve a pull request
      • Merge a pull request or patch
    • Build the code and run the tests
    • Create a Storm distribution (packaging)
  • Best practices
    • Testing
    • Version Changes
  • Tools
    • Source code repositories (git)
    • Issue tracking (JIRA)
  • Questions?


If you are reading this document then you are interested in contributing to the Storm project -- many thanks for that! All contributions are welcome: ideas, documentation, code, patches, bug reports, feature requests, etc. And you do not need to be a programmer to speak up.


This section explains how to perform common activities such as reporting a bug or merging a pull request.

Report a bug

To report a bug you should open an issue in our issue tracker that summarizes the bug. Set the form field “Issue type” to “Bug”. If you have not used the issue tracker before you will need to register an account (free), log in, and then click on the blue “Create Issue” button in the top navigation bar.

In order to help us understand and fix the bug it would be great if you could provide us with:

  1. The steps to reproduce the bug. This includes information about e.g. the Storm version you were using.
  2. The expected behavior.
  3. The actual, incorrect behavior.

Feel free to search the issue tracker for existing issues (aka tickets) that already describe the problem; if there is such a ticket please add your information as a comment.

If you want to provide a patch along with your bug report: That is great! In this case please send us a pull request as described in section Create a pull request below. You can also opt to attach a patch file to the issue ticket, but we prefer pull requests because they are easier to work with.

Request a new feature

To request a new feature you should open an issue in our issue tracker and summarize the desired functionality. Set the form field “Issue type” to “New feature”. If you have not used the issue tracker before you will need to register an account (free), log in, and then click on the blue “Create Issue” button in the top navigation bar.

You can also opt to send a message to the Storm Users mailing list.

Contribute code

Before you set out to contribute code we recommend that you familiarize yourself with the Storm codebase, notably by reading through the Implementation documentation.

If you are interested in contributing code to Storm but do not know where to begin: In this case you should browse our issue tracker for open issues and tasks. You may want to start with beginner-friendly, easier issues (newbie issues and trivial issues) because they require learning about only an isolated portion of the codebase and are a relatively small amount of work.

Please use idiomatic Clojure style, as explained in this Clojure style guide. Another useful reference is the Clojure Library Coding Standards. Perhaps the most important is consistenly writing a clear docstring for functions, explaining the return value and arguments. As of this writing, the Storm codebase would benefit from various style improvements.

Contributions to the Storm codebase should be sent as GitHub pull requests. See section Create a pull request below for details. If there is any problem with the pull request we can iterate on it using the commenting features of GitHub.

  • For small patches, feel free to submit pull requests directly for those patches.

  • For larger code contributions, please use the following process. The idea behind this process is to prevent any wasted work and catch design issues early on.

    1. Open an issue on our issue tracker if a similar issue does not exist already. If a similar issue does exist, then you may consider participating in the work on the existing issue.
    2. Comment on the issue with your plan for implementing the issue. Explain what pieces of the codebase you are going to touch and how everything is going to fit together.
    3. Storm committers will iterate with you on the design to make sure you are on the right track.
    4. Implement your issue, create a pull request (see below), and iterate from there.


Unit tests and Integration tests are an essential part of code contributions.

To mark a Java test as a Java integration test, add the annotation @IntegrationTest to the test class definition or test method. Make sure the test is a JUnit 5 test. Java integration tests can be in the same package as Java unit tests.

    public class MyIntegrationTest {

To mark a Clojure test as Clojure integration test, the test source must be located in a package with name prefixed by integration.

For example, the test test/clj/org.apache.storm.drpc_test.clj is considered a clojure unit test, whereas test/clj/ is considered a clojure integration test.

Please refer to section Build the code and run the tests for how to run integration tests, and the info on the build phase each test runs.

Contribute documentation

Documentation contributions are very welcome!

You can contribute documentation by pull request, using the same process as code contribution. Release specific documentation can be found in docs/ in this repository. Documentation not specific to a release, e.g. announcements, is found in the storm-site repository. If you'd like to build or test the website, refer to the in the storm-site repository.

Pull requests

Create a pull request

Pull requests should be done against the read-only git repository at

Take a look at Creating a pull request. In a nutshell you need to:

  1. Fork the Storm GitHub repository at to your personal GitHub account. See Fork a repo for detailed instructions.
  2. Commit any changes to your fork.
  3. Send a pull request to the Storm GitHub repository that you forked in step 1. If your pull request is related to an existing Storm JIRA ticket -- for instance, because you reported a bug report via JIRA earlier -- then prefix the title of your pull request with the corresponding JIRA ticket number (e.g. STORM-123: ...).

You may want to read Syncing a fork for instructions on how to keep your fork up to date with the latest changes of the upstream (official) storm repository.

Approve a pull request

BYLAWS describes the condition of approval for code / non-code change.

Please refer Approvals -> Actions section for more details.

Merge a pull request or patch

This section applies to committers only.

Important: A pull request must first be properly approved before you are allowed to merge it.

Committers that are integrating patches or pull requests should use the official Apache repository at

Via Github

You can use the Gitbox account linking utility to link your Apache and Github accounts. This will allow you to merge pull requests using Github's UI.

Via your terminal

To pull in a merge request you should generally follow the command line instructions sent out by GitHub.

  1. Go to your local copy of the Apache git repo, switch to the master branch, and make sure it is up to date.

     $ git checkout master
     $ git fetch origin
     $ git merge origin/master
  2. Create a local branch for integrating and testing the pull request. You may want to name the branch according to the Storm JIRA ticket associated with the pull request (example: STORM-1234).

     $ git checkout -b <local_test_branch>  # e.g. git checkout -b STORM-1234
  3. Merge the pull request into your local test branch.

     $ git pull <remote_repo_url> <remote_branch>

    You can use ./dev-tools/ <pull-number> to produce the above command most of the time.

  4. Run any sanity tests that you think are needed.

  5. Once you are confident that everything is ok, you can merge your local test branch into your local master branch, and push the changes back to the official Apache repo.

     # Pull request looks ok, change log was updated, etc.  We are ready for pushing.
     $ git checkout master
     $ git merge <local_test_branch>  # e.g. git merge STORM-1234
     # At this point our local master branch is ready, so now we will push the changes
     # to the official Apache repo.  Note: The read-only mirror on GitHub will be updated
     # automatically a short while after you have pushed to the Apache repo.
     $ git push origin master
  6. The last step is updating the corresponding JIRA ticket. Go to JIRA and resolve the ticket. Be sure to set the Fix Version/s field to the version you pushed your changes to. It is usually good practice to thank the author of the pull request for their contribution if you have not done so already.

Build the code and run the tests


In order to build storm you need python, ruby and nodejs. In order to avoid an overfull page we don‘t provide platform/OS specific installation instructions for those here. Please refer to you platform’s/OS' documentation for support.

The ruby package manager rvm and nodejs package manager nvm are for convenience and are used in the tests which run on travis. They can be installed using curl -L | bash -s stable --autolibs=enabled && source ~/.profile (see the rvm installation instructions for details) and wget -qO- | bash && source ~/.bashrc (see the nvm installation instructions for details).

With rvm and nvm installed you can run

rvm use 2.4.2 --install
nvm install 8.9.3
nvm use 8.9.3

in order to get started as fast as possible. Users can still install a specific version of ruby and/or node manually.

You will also need the mock Python testing library (as well as Python 2.7.x and Python 3.x). With pip installed you can run

pip install mock


The following commands must be run from the top-level directory.

mvn clean install

If you wish to skip the unit tests you can do this by adding -DskipTests to the command line.

If you wish to skip the examples and external modules, you can do this by adding -P '!examples,!externals' to the command line.

In case you modified storm.thrift, you have to regenerate thrift code as java and python code before compiling whole project.

cd storm-client/src


Tests are separated in two groups, Unit tests, and Integration tests. Java unit tests, Clojure unit tests, and Clojure integration tests (for reasons inherent to the clojure-maven-plugin) run in the maven test phase. Java integration tests run in the maven integration-test or verify phases.

To run Clojure and Java unit tests but no integration tests execute the command

mvn test

Integration tests require that you activate the profile integration-test and that you specify the maven-failsafe-plugin in the module pom file.

To run all Java and Clojure integration tests but no unit tests execute one of the commands

mvn -P integration-tests-only,examples,externals verify
mvn -P integration-tests-only,examples,externals integration-test

To run all unit tests plus Clojure integration tests but no Java integration tests execute the command

mvn -P all-tests,examples,externals test

To run all unit tests and all integration tests execute one of the commands

mvn -P all-tests,examples,externals verify
mvn -P all-tests,examples,externals integration-test

You can also run tests selectively with -Dtest=<test_name>. This works for both clojure and junit tests.

Unfortunately you might experience failures in clojure tests which are wrapped in the maven-clojure-plugin and thus doesn‘t provide too much useful output at first sight - you might end up with a maven test failure with an error message as unhelpful as Clojure failed.. In this case it’s recommended to look into target/test-reports of the failed project to see what actual tests have failed or scroll through the maven output looking for obvious issues like missing binaries.

By default integration tests are not run in the test phase. To run Java and Clojure integration tests you must enable the profile integration-tests-only, or all-tests.

Listing dependency licenses

You can generate a list of dependencies and their licenses by running mvn license:aggregate-add-third-party@generate-and-check-licenses -Dlicense.skipAggregateAddThirdParty=false in the project root. The list will be put in DEPENDENCY_LICENSES.

The license aggregation plugin will use the license listed in a dependency's POM. If the license is missing, or incomplete (e.g. due to multiple licenses), you can override the license by describing the dependency in the file in the project root.

Auditing licenses for LICENSE/NOTICE

The LICENSE and NOTICE files contain licenses and notices for source distribution content. The LICENSE-binary and NOTICE-binary apply to the binary distributions.

When auditing the binary LICENSE-binary and NOTICE-binary, there are a couple of helper scripts available in dev-tools. collect_license_files can create an aggregate NOTICE from the libraries in an extracted distribution. The aggregate NOTICE should be adjusted to remove Storm notices and duplicates, and added to the NOTICE-binary.

The dev-tools/ script will check that LICENSE-binary and DEPENDENCY_LICENSES are up to date. Regenerating DEPENDENCY_LICENSES simply requires rerunning the license plugin (see above). LICENSE-binary must be updated manually. The script will check that the dependencies included in a storm-dist/binary build are present in LICENSE-binary, and that no other dependencies are listed. Any additional or missing dependencies are printed to console, and can be added to LICENSE-binary manually. There will likely be an entry for them in DEPENDENCY_LICENSES that can be copy-pasted to LICENSE-binary.

You can download the dependency licenses by running mvn package -Dlicense.skipAggregateDownloadLicenses=false -DskipTests in the project root. This will put the licenses in target/generated-resources. Keep an eye on the Maven output, as some dependencies may not have licenses configured correctly. These will have to be downloaded manually.

Create a Storm distribution (packaging)

You can create a distribution (like what you can download from Apache) as follows. Note that the instructions below do not use the Maven release plugin because creating an official release is the task of our release manager.

# First, build the code.
$ mvn clean install # you may skip tests with `-DskipTests=true` to save time

# Create the binary distribution.
$ cd storm-dist/binary && mvn package

The last command will create Storm binaries at:


including corresponding *.asc digital signature files.

After running mvn package you may be asked to enter your GPG/PGP credentials (once for each binary file, in fact). This happens because the packaging step will create *.asc digital signatures for all the binaries, and in the workflow above your GPG private key will be used to create those signatures.

You can verify whether the digital signatures match their corresponding files:

# Example: Verify the signature of the `.tar.gz` binary.
$ gpg --verify storm-dist/binary/target/apache-storm-<version>.tar.gz.asc

Best practices


Tests should never rely on timing in order to pass. Storm can properly test functionality that depends on time by simulating time, which means we do not have to worry about e.g. random delays failing our tests indeterministically.

If you are testing topologies that do not do full tuple acking, then you should be testing using the “tracked topologies” utilities in org.apache.storm.testing.clj. For example, test-acking (around line 213) tests the acking system in Storm using tracked topologies. Here, the key is the tracked-wait function: it will only return when both that many tuples have been emitted by the spouts and the topology is idle (i.e. no tuples have been emitted nor will be emitted without further input). Note that you should not use tracked topologies for topologies that have tick tuples.

Version Changes

An easy way to change versions across all pom files, for example from 1.0.0-SNAPSHOT to 1.0.0, is with the maven versions plugin.

mvn versions:set #This prompts for a new version
mvn versions:commit

[Plugin Documentation] (


Source code repositories (git)

The source code of Storm is managed via git. For a number of reasons there is more than one git repository associated with Storm.

  • Committers only: is the official and authoritative git repository for Storm, managed under the umbrella of the Apache Software Foundation. Only official Storm committers will interact with this repository. When you push the first time to this repository git will prompt you for your username and password. Use your Apache user ID and password, i.e. the credentials you configured via after you were onboarded as a committer.
  • Everybody else: is a read-only mirror of the official git repository. If you are not a Storm committer (most people) this is the repository you should work against. See Development workflow above on how you can create a pull request, for instance.

An automated bot (called ASF GitHub Bot in Storm JIRA) runs periodically to merge changes in the official Apache repo to the read-only GitHub mirror repository, and to merge comments in GitHub pull requests to the Storm JIRA.

Issue tracking (JIRA)

Issue tracking includes tasks such as reporting bugs, requesting and collaborating on new features, and administrative activities for release management. As an Apache software project we use JIRA as our issue tracking tool.

The Storm JIRA is available at:

If you do not have a JIRA account yet, then you can create one via the link above (registration is free).

The storm codebase is available at Codota where you can semantically search it like in an IDE (e.g. find usages for a method).


If you have any questions after reading this document, then please reach out to us via the Storm Developers mailing list.

And of course we also welcome any contributions to improve the information in this document!