Airflow and Spark: Running Spark jobs on Airflow (Docker-based solution)

Here in this repository, we have designed a simple ETL process that extract data from an API and we are transforming this data using Spark and loading this data into an AWS S3 bucket. We running this batch processes using Airflow by Spark job submit Operator in Airflow. All the processes described here are happening on a Docker containers. You can look at this repository if you are interested in local deployment as opposed to Docker-based solution.

alt text

Things to do;

Clone the Github repository
Build the Spark and the Airflow image
Create your dags, logs, plugins folder
Create your environment variable
Start and run the Spark and Airflow containers
Run your Spark jobs to confirm if the Spark job completed successfully before moving it to Airflow
Design the Airflow DAG to trigger and schedule the Spark jobs.

Clone the Github repository.

git clone https://github.com/yTek01/docker-spark-airflow.git

Build the Spark image.

docker build -f Dockerfile.Spark . -t spark-air

Build the Airflow image.

docker build -f Dockerfile.Airflow . -t airflow-spark

Create your dags, logs, plugins folder.

mkdir ./dags ./logs ./plugins
echo -e "AIRFLOW_UID=$(id -u)\nAIRFLOW_GID=0" > .env

Your environment variable would look like this.

AIRFLOW_UID=33333
AIRFLOW_GID=0
AWS_ACCESS_KEY=XXXXXXXXXXXXXXXXXXXX
AWS_SECRET_KEY=XXXXXXXXXXXXXXXXXXXX

Start and run the Spark and Airflow containers.

docker-compose -f docker-compose.Spark.yaml -f docker-compose.Airflow.yaml up -d

When all the services all started successfully, now go to http://localhost:8080/ to check that Airflow has started successfully, and http://localhost:8090/ that Spark is up and running.

Run your Spark jobs to confirm if the Spark job completed successfully before moving it to Airflow.

docker exec -it <Spark-Worker-Contianer-name> \
    spark-submit --master spark://XXXXXXXXXXXXXX:7077 \
    spark_etl_script_docker.py

If all is fine with the setup, i.e. the Spark job completed successfully, then move forward to scheduling the Spark job on Airflow.

yTek01 / docker-spark-airflow

readme