This is a repository for building Docker container of Apache Airflow (incubating).

• For understanding & knowing more about Airflow, please follow curated list of resources.
• Similarly, for Docker follow curated list of resources.

Image	Pulls	Tags
abhioncbr/docker-airflow		tags

• Airflow version: Notation for representing version XX.YY.ZZ
• Execution Mode: standalone(simple container for exploration purpose, based on sqlite as airflow metadata db & SequentialExecutor ) or prod(single node based, LocalExecutor amd mysql as airflow metadata db) and cluster (for distributed production long run use-cases, container runs as either server or worker )
• Backend database: standalone- Sqlite, prod & cluster- Mysql
• Scheduler: standalone- Sequential, prod- LocalExecutor and Cluster- Celery
• Task queue: cluster- Redis
• Log location: local file system (Default) or AWS S3 (through entrypoint-s3.sh)
• User authentication: Password based & support for multiple users with superuser privilege.
• Code enhancement: password based multiple users supporting super-user(can see all dags of all owner) feature. Currently, Airflow is working on the password based multi user feature.
• Other features: support for google cloud platform packages in container.

• airflow portal port: 2222
• airflow celery flower: 5555
• redis port: 6379
• log files exchange port: 8793

• In server container: redis, airflow webserver & scheduler is running.
• In worker container: airflow worker & celery flower ui service is running.

• DockerFile uses airflow-version as a build-arg.
• build image, if you want to do some customization -

     docker build -t abhioncbr/docker-airflow:$IMAGE_VERSION --build-arg AIRFLOW_VERSION=$AIRFLOW_VERSION
                --build-arg AIRFLOW_PATCH_VERSION=$AIRFLOW_PATCH_VERSION -f ~/docker-airflow/docker-files/DockerFile .

  • Arg IMAGE_VERSION value should be airflow version for example, 1.10.3 or 1.10.2
  • Arg AIRFLOW_PATCH_VERSION value should be the major release version of airflow for example for 1.10.2 it should be 1.10.

• Simplest way for exploration purpose, using Kitematic(Run containers through a simple, yet powerful graphical user interface.)

  • Search abhioncbr/docker-airflow Image on docker-hub

  • Start a container through Kitematic UI.

• General commands -

  • starting airflow image as a airflow-standalone container in a standalone mode-

    docker run --net=host -p 2222:2222 --name=airflow-standalone abhioncbr/airflow-XX.YY.ZZ -m=standalone &

  • Starting airflow image as a airflow-server container in a cluster mode-

    docker run --net=host -p 2222:2222 -p 6379:6379 --name=airflow-server \
    abhioncbr/airflow-XX.YY.ZZ -m=cluster -t=server -d=mysql:https://user:password@host:3306/db-name &

  • Starting airflow image as a airflow-worker container in a cluster mode-

    docker run --net=host -p 5555:5555 -p 8739:8739 --name=airflow-worker \
    abhioncbr/airflow-XX.YY.ZZ -m=cluster -t=worker -d=mysql:https://user:password@host:3306/db-name -r=redis:https://<airflow-server-host>:6379/0 &

• In Mac using docker for mac -

  • Standalone Mode - starting airflow image in a standalone mode & mounting dags, code-artifacts & logs folder to host machine -

    docker run -p 2222:2222 --name=airflow-standalone \
    -v ~/airflow-data/code-artifacts:/code-artifacts \
    -v ~/airflow-data/logs:/usr/local/airflow/logs \
    -v ~/airflow-data/dags:/usr/local/airflow/dags \
    abhioncbr/airflow-XX.YY.ZZ -m=standalone &

  • Cluster Mode

    • starting airflow image as a server container & mounting dags, code-artifacts & logs folder to host machine -

      docker run -p 2222:2222 -p 6379:6379 --name=airflow-server \
      -v ~/airflow-data/code-artifacts:/code-artifacts \
      -v ~/airflow-data/logs:/usr/local/airflow/logs \
      -v ~/airflow-data/dags:/usr/local/airflow/dags \
      abhioncbr/airflow-XX.YY.ZZ \
      -m=cluster -t=server -d=mysql:https://user:[email protected]:3306:3306/<airflow-db-name> &

    • starting airflow image as a worker container & mounting dags, code-artifacts & logs folder to host machine -

      docker run -p 5555:5555 -p 8739:8739 --name=airflow-worker \
      -v ~/airflow-data/code-artifacts:/code-artifacts \
      -v ~/airflow-data/logs:/usr/local/airflow/logs \
      -v ~/airflow-data/dags:/usr/local/airflow/dags \
      abhioncbr/airflow-XX.YY.ZZ \
      -m=cluster -t=worker -d=mysql:https://user:[email protected]:3306:3306/<airflow-db-name> -r=redis:https://host.docker.internal:6379/0 &   

  

• As mentioned above, docker image of airflow can be leveraged to run in complete distributed run
  • single docker-airflow container in server mode for serving the UI of the airflow, redis for celery task & scheduler.
  • multiple docker-airflow containers in worker mode for executing tasks using celery executor.
  • centralised airflow metadata database.
• Image below depicts the docker-airflow distributed platform: