The Debezium community welcomes anyone that wants to help out in any way, whether that includes reporting problems, helping with documentation, or contributing code changes to fix bugs, add tests, or implement new features. This document outlines the basic steps required to work with and contribute to the Debezium codebase.
Before doing any code changes, it's a good idea to reach out to us, so as to make sure there's a general consencus on the proposed change and the implementation strategy. You can reach us here:
- User chat
- Developers chat - Only for internal development subjects
- Google Group
- Jira
The following software is required to work with the Debezium codebase and build it locally:
- Git 2.2.1 or later
- JDK 11 or later, e.g. OpenJDK
- Apache Maven 3.8.4
- Docker Engine or Docker Desktop 1.9 or later
See the links above for installation instructions on your platform. You can verify the versions are installed and running:
$ git --version
$ javac -version
$ mvn -version
$ docker --version
Debezium uses GitHub for its primary code repository and for pull-requests, so if you don't already have a GitHub account you'll need to join.
Debezium uses the Red Hat Jira instance for issue tracking. If you don't have an account yet, you can create one for free by clicking "Log In" and then signing up with Red Hat's SSO service.
Go to the Debezium repository and press the "Fork" button near the upper right corner of the page. When finished, you will have your own "fork" at https://github.com/<your-username>/debezium, and this is the repository to which you will upload your proposed changes and create pull requests. For details, see the GitHub documentation.
At a terminal, go to the directory in which you want to place a local clone of the Debezium repository, and run the following commands to use HTTPS authentication:
$ git clone https://github.com/<your-username>/debezium.git
If you prefer to use SSH and have uploaded your public key to your GitHub account, you can instead use SSH:
$ git clone git@github.com:<your-username>/debezium.git
This will create a debezium directory, so change into that directory:
$ cd debezium
This repository knows about your fork, but it doesn't yet know about the official or "upstream" Debezium repository. Run the following commands:
$ git remote add upstream https://github.com/debezium/debezium.git
$ git fetch upstream
$ git branch --set-upstream-to=upstream/main main
Now, when you check the status using Git, it will compare your local repository to the upstream repository.
You will frequently need to get all the of the changes that are made to the upstream repository, and you can do this with these commands:
$ git fetch upstream
$ git pull upstream main
The first command fetches all changes on all branches, while the second actually updates your local main branch with the latest commits from the upstream repository.
To build the source code locally, checkout and update the main branch:
$ git checkout main
$ git pull upstream main
Then use Maven to compile everything, run all unit and integration tests, build all artifacts, and install all JAR, ZIP, and TAR files into your local Maven repository:
$ mvn clean install -Passembly
If you want to skip the integration tests (e.g., if you don't have Docker installed) or the unit tests, you can add -DskipITs and/or -DskipTests to that command:
$ mvn clean install -Passembly -DskipITs -DskipTests
A number of the modules use Docker during their integration tests to run a database. During development it's often desirable to start the Docker container and leave it running so that you can compile/run/debug tests repeatedly from your IDE. To do this, simply go into one of the modules (e.g., cd debezium-connector-mysql) and run the following command:
$ mvn docker:build docker:start
This will first force the build to create a new Docker image for the database container, and then will start a container named "database". You can then run any integration tests from your IDE, though all of our integration tests expect the database connection information to be passed in as system variables (like our Maven build does). For example, the MySQL connector integration tests expect something like -Ddatabase.hostname=localhost -Ddatabase.port=3306 to be passed as arguments to your test.
When your testing is complete, you can stop the Docker container by running:
$ mvn docker:stop
or the following Docker commands:
$ docker stop database; docker rm database
Everything the community does with the codebase -- fixing bugs, adding features, making improvements, adding tests, etc. -- should be described by an issue in our Jira project. If no such issue exists for what you want to do, please create an issue with a meaningful and easy-to-understand description. If you are going to work on a specific issue and it's your first contribution, please add a short comment to the issue, so other people know you're working on it. If you are contributing repeatedly, ask in our chat room for developers for the 'developer' Jira role so you can assign issues to yourself.
Before you make any changes, be sure to switch to the main branch and pull the latest commits on the main branch from the upstream repository. Also, it's probably good to run a build and verify all tests pass before you make any changes.
$ git checkout main
$ git pull upstream main
$ mvn clean install
Once everything builds, create a topic branch named appropriately (we recommend using the issue number, such as DBZ-1234):
$ git checkout -b DBZ-1234
This branch exists locally and it is there you should make all of your proposed changes for the issue. As you'll soon see, it will ultimately correspond to a single pull request that the Debezium committers will review and merge (or reject) as a whole. (Some issues are big enough that you may want to make several separate but incremental sets of changes. In that case, you can create subsequent topic branches for the same issue by appending a short suffix to the branch name.)
Your changes should include changes to existing tests or additional unit and/or integration tests that verify your changes work. We recommend frequently running related unit tests (in your IDE or using Maven) to make sure your changes didn't break anything else, and that you also periodically run a complete build using Maven to make sure that everything still works:
$ mvn clean install
Feel free to commit your changes locally as often as you'd like, though we generally prefer that each commit represent a complete and atomic change to the code. Often, this means that most issues will be addressed with a single commit in a single pull-request, but other more complex issues might be better served with a few commits that each make separate but atomic changes. (Some developers prefer to commit frequently and to ammend their first commit with additional changes. Other developers like to make multiple commits and to then squash them. How you do this is up to you. However, never change, squash, or ammend a commit that appears in the history of the upstream repository.) When in doubt, use a few separate atomic commits; if the Debezium reviewers think they should be squashed, they'll let you know when they review your pull request.
Committing is as simple as:
$ git commit .
which should then pop up an editor of your choice in which you should place a good commit message. We do expect that all commit messages begin with a line starting with the Jira issue and ending with a short phrase that summarizes what changed in the commit. For example:
DBZ-1234 Expanded the MySQL integration test and correct a unit test.
If that phrase is not sufficient to explain your changes, then the first line should be followed by a blank line and one or more paragraphs with additional details. For example:
DBZ-1235 Added support for ingesting data from PostgreSQL.
The new ingest library supports PostgreSQL 9.4 or later. It requires a small plugin to be installed
on the database server, and the database to be configured so that the ingest component can connect
to the database and use logical decoding to read the transaction log. Several new unit tests and one
integration test were added.
As an exception, commits for trivial documentation changes which don't warrant the creation of a Jira issue can be prefixed with [docs], for example:
[docs] Typo fix in Postgres connector documentation
The following commit message prefixes are reserved for internal usage by the Debezium CI infrastructure and should not be used in manually created commits: [release], [jenkins-jobs], [maven-release-plugin], and [ci].
This project utilizes a set of code style rules that are automatically applied by the build process. There are two ways in which you can apply these rules while contributing:
-
If your IDE supports importing an Eclipse-based formatter file, navigate to your IDE's code format section and import the file
support/ide-configs/src/main/resources/eclipse/debezium-formatter.xml. It's recommended that when you import these styles, you may want to make sure these are only applicable only to the current project. -
If your IDE does not support importing the Eclipse-based formatter file or you'd rather tidy up the formatting after making your changes locally, you can run a project build to make sure that all code changes adhere to the project's desired style. Instructions on how to run a build locally are provided below.
-
With the command
mvn process-sourcesthe code style rules can be applied automatically.
In the event that a pull request is submitted with code style violations, continuous integration will fail the pull request build.
To fix pull requests with code style violations, simply run the project's build locally and allow the automatic formatting happen. Once the build completes, you will have some local repository files modified to fix the coding style which can be amended on your pull request. Once the pull request is synchronized with the formatting changes, CI will rerun the build.
To run the build, navigate to the project's root directory and run:
$ mvn clean install
It might be useful to simply run a validate check against the code instead of automatically applying code style changes. If you want to simply run validation, navigate to the project's root directory and run:
$ mvn clean install -Dformat.formatter.goal=validate -Dformat.imports.goal=check
Please note that when running validate checks, the build will stop as soon as it encounters its first violation. This means it is necessary to run the build multiple times until no violations are detected.
If its been more than a day or so since you created your topic branch, we recommend rebasing your topic branch on the latest main branch. This requires switching to the main branch, pulling the latest changes, switching back to your topic branch, and rebasing:
$ git checkout main
$ git pull upstream main
$ git checkout DBZ-1234
$ git rebase main
If your changes are compatible with the latest changes on main, this will complete and there's nothing else to do. However, if your changes affect the same files/lines as other changes have since been merged into the main branch, then your changes conflict with the other recent changes on main, and you will have to resolve them. The git output will actually tell you you need to do (e.g., fix a particular file, stage the file, and then run git rebase --continue), but if you have questions consult Git or GitHub documentation or spend some time reading about Git rebase conflicts on the Internet.
When adding new features such as e.g. a connector or configuration options, they must be documented accordingly in the Debezium reference documentation. The same applies when changing existing behaviors, e.g. type mappings, removing options etc.
The documentation is written using AsciiDoc/Antora and can be found in the Debezium source code repository. Any documentation update should be part of the pull request you submit for the code change.
Once you're finished making your changes, your topic branch should have your commit(s) and you should have verified that your branch builds successfully. At this point, you can shared your proposed changes and create a pull request. To do this, first push your topic branch (and its commits) to your fork repository (called origin) on GitHub:
$ git push origin DBZ-1234
Then, in a browser go to your forked repository, and you should see a small section near the top of the page with a button labeled "Contribute". GitHub recognized that you pushed a new topic branch to your fork of the upstream repository, and it knows you probably want to create a pull request with those changes. Click on the button, and a button "Open pull request" will apper. Click it and GitHub will present you the "Comparing changes" page, where you can view all changes that you are about to submit. With all revised, click in "Create pull request" and a short form will be given, that you should fill out with information about your pull request. The title should start with the Jira issue and end with a short phrase that summarizes the changes included in the pull request. (If the pull request contains a single commit, GitHub will automatically prepopulate the title and description fields from the commit message.)
When completed, press the "Create" button and copy the URL to the new pull request. Go to the corresponding Jira issue and record the pull request by pasting the URL into the "Pull request" field. (Be sure to not overwrite any URLs that were already in this field; this is how a single issue is bound to multiple pull requests.) Also, please add a Jira comment with a clear description of what you changed. You might even use the commit message (except for the first line).
At this point, you can switch to another issue and another topic branch. The Debezium committers will be notified of your new pull request, and will review it in short order. They may ask questions or make remarks using line notes or comments on the pull request. (By default, GitHub will send you an email notification of such changes, although you can control this via your GitHub preferences.)
If the reviewers ask you to make additional changes, simply switch to your topic branch for that pull request:
$ git checkout DBZ-1234
and then make the changes on that branch and either add a new commit or ammend your previous commits. When you've addressed the reviewers' concerns, push your changes to your origin repository:
$ git push origin DBZ-1234
GitHub will automatically update the pull request with your latest changes, but we ask that you go to the pull request and add a comment summarizing what you did. This process may continue until the reviewers are satisfied.
By the way, please don't take offense if the reviewers ask you to make additional changes, even if you think those changes are minor. The reviewers have a broad understanding of the codebase, and their job is to ensure the code remains as uniform as possible, is of sufficient quality, and is thoroughly tested. When they believe your pull request has those attributes, they will merge your pull request into the official upstream repository.
Once your pull request has been merged, feel free to delete your topic branch both in your local repository:
$ git branch -d DBZ-1234
and in your fork:
$ git push origin :DBZ-1234
(This last command is a bit strange, but it basically is pushing an empty branch (the space before the : character) to the named branch. Pushing an empty branch is the same thing as removing it.)
The project currently builds its jobs in two environments:
- GitHub Actions for pull requests: https://github.com/debezium/debezium/actions
- Tests run only against the current version of each supported database
- Jenkins CI for tests matrix, deployment, release, etc - http://ci.hibernate.org/view/Debezium/
- Test run against all database versions supported by the individual connectors
- Test Kafka versions
- Deploy and release
Here's a quick check list for a good pull request (PR):
- Discussed and approved on Jira, chat or the mailing list
- A Jira associated with your PR (include the Jira issue number in commit comment)
- One feature/change per PR
- No changes to code not directly related to your change (e.g. no formatting changes or refactoring to existing code, if you want to refactor/improve existing code that's a separate discussion and separate Jira issue)
- New/changed features have been documented
- A full build completes successfully
- Do a rebase on upstream
main
- No code changes without PR (unless it's trivial typo fixes and the like; just commit those)
- Don't merge your own PRs, ensure four eyes principle (unless it's trivial changes, or a reviewer suggested to merge yourself after applying some suggestions)
- Ensure there are no "Octocat" commits
- During reviews, focus on the relevant parts (APIs, semantics, functional behavior, performance aspects, etc); de-focus on style and formatting (the latter is automated anyways)
- Don't do force pushes to master/main branches (it's disabled for all repos)
- Always apply PRs via rebasing instead of merges (there should be a linear commit history)
- Optional: squash commits into one if there's lots of forth and back on a PR which isn't relevant for the history in the long run