A small boilerplate/project with docker for data science.

cp config/.env.jupyter.example .env.jupyter
cp config/.env.minio.example .env.minio
cp config/.env.postgres.example .env.postgres
cp config/.env.airflow.example .env.airflow
cp config/.env.database.example .env.database

Provide appropriate values in the .env files and then run

docker-compose up -d

if you want to persist jupyter settings, make sure to commit the container before taking it down. So

docker commit jupyter
# and then
docker-compose down

Otherwise you can simply stick to docker-compose start and stop.

Services:

• jupyterlab: Jupyter notebooks and jupyter lab where you can do fancy stuff (localhost:8888)
• minio: A key value file store like aws (localhost:9001)
• postgres: Database store
• metabase: Cool data science stuff
• superset: Another cool visualisations service
• airflow: Scheduler and task runner (localhost:8080)

Structure:

• config: Contains environment variables, keys, secrets etc
• jupyter: Contains notebooks
• dags: Task runners

Data folders:

• data: Contains data which is mounted on to minio
• db_data: Postgres database persistant volumne
• metabase_data: Metabase data persistant volume

Inspired by data-science-stack