A JVM SDK for Apache Airflow. You can use any JVM-compatible language to write workflow bundles, and have Airflow consume the result.

The SDK and execution-time logic is implemented in Kotlin. An example is bundled showing how the SDK can be used in Java.

./gradlew build

./gradlew dokkaGenerate

This uses Dokka to build documentation of the Java SDK. This generates both an HTML representation and Javadoc.

• Put the DAG with stub tasks to somewhere Airflow can find.

• Ensure the java command is available in the same environment the Airflow task worker is in.

• Package the example and its dependencies into JARs in ./example/build/install/example/lib

  ./gradlew :example:installDist

• Configure Airflow to route tasks in the java queue to be run with Java:

  export AIRFLOW__SDK__COORDINATORS='{
    "java": {
      "classpath": "airflow.sdk.coordinators.java.JavaCoordinator",
      "kwargs": {"jars_root": ["/opt/airflow/java-sdk/example/build/install/example/lib"]}
    }
  }'
  export AIRFLOW__SDK__QUEUE_TO_COORDINATOR='{"java": "java"}'

• Ensure the Connection and Variable needed by the example DAG are available:

  export AIRFLOW_CONN_TEST_HTTP='{
      "conn_type": "http",
      "login": "user",
      "password": "pass",
      "host": "example.com",
      "port": 1234,
      "extra": {"param1": "val1", "param2": "val2"}
  }'
  export AIRFLOW_VAR_MY_VARIABLE=123

The SDK is published to Maven Central via the ASF Nexus staging repository. The full release process follows the ASF Maven publishing guide.

• An ASF committer account with access to repository.apache.org.
• A GPG key that has been added to the project KEYS file and uploaded to a public key server.

Edit gradle.properties and set the version for this release:

sdkVersion=1.0.0

Commit the change and push it to the release branch.

Before touching any remote repository, publish to your local Maven cache and inspect the generated POM:

./gradlew :sdk:publishToMavenLocal
cat ~/.m2/repository/org/apache/airflow/airflow-sdk/*/airflow-sdk-*.pom

Check that the coordinates, description, license, SCM, and organization fields look correct.

To test the full publish flow without touching ASF infrastructure, override the repository URL to a local directory

./gradlew :sdk:publish -PmavenUrl=file:///tmp/local-maven-repo -PskipSigning
ls /tmp/local-maven-repo/org/apache/airflow/airflow-sdk/

NOTE: Signing is not required since nothing goes to Maven Central. If you want to test signing, set the GPG private key and passphrase as described in the next section, and remove -PskipSigning from the above command.

Store the credentials in ~/.gradle/gradle.properties so they are not exposed in your shell history:

mavenUsername=your-asf-nexus-token-username
mavenPassword=your-asf-nexus-token-password
signing.password=your-gpg-key-passphrase

Then run the publish task.

./gradlew :sdk:publish -P"signing.key=$(gpg --armor --export-secret-keys your-gpg-key-fingerprint)"

NOTE: The signing key is supplied through the command line since it contains newlines, which does not work well in a Gradle properties file.

NOTE: You can also use the following environment variables to provide the credentials instead: ASF_NEXUS_USERNAME, ASF_NEXUS_PASSWORD, SIGNING_KEY, and SIGNING_PASSWORD. This is especially useful on e.g. CI.

The user uses the SDK to implement a Java application that implements task methods, and metadata on which DAG and task each method should be used for.

When the Airflow Supervisor identifies a task should be run with Java, it launches the Java application as a subprocess. The Java application accepts flags --comm and --logs from the command line to identify TCP sockets it should connect to, and communicates with the Supervisor through these channels during execution.

1. On connection, the Supervisor immediately sends a StartupDetails message through the comm socket.
2. The Java application finds and executes the relevant method.
3. During execution, the Java application uses the comm socket to retrieve information (e.g. Variable) from, and send data (e.g. XCom) to Airflow.
4. The Java application informs the comm socket to tell the Supervisor the task's terminal state.
5. The Java application exits.

During the Java application's lifetime, it also sends log messages generated by the SDK (not user code) through the logs socket, so the Supervisor can append them to Airflow logs.

Communication uses the same formats as the Python-based processes.

See Architectural Design Records in the adr directory to learn more.