This repository contains Dockerfile of apache-airflow for Docker's automated build published to the public Docker Hub Registry.
- Based on Python (3.7-slim-buster) official Image python:3.7-slim-buster and uses the official Postgres as backend and Redis as queue
- Install Docker
- Install Docker Compose
- Following the Airflow release from Python Package Index
Pull the image from the Docker repository.
docker pull puckel/docker-airflow
Optionally install Extra Airflow Packages and/or python dependencies at build time :
docker build --rm --build-arg AIRFLOW_DEPS="datadog,dask" -t puckel/docker-airflow .
docker build --rm --build-arg PYTHON_DEPS="flask_oauthlib>=0.9" -t puckel/docker-airflow .
or combined
docker build --rm --build-arg AIRFLOW_DEPS="datadog,dask" --build-arg PYTHON_DEPS="flask_oauthlib>=0.9" -t puckel/docker-airflow .
Don't forget to update the airflow images in the docker-compose files to puckel/docker-airflow:latest.
By default, docker-airflow runs Airflow with SequentialExecutor :
docker run -d -p 8080:8080 puckel/docker-airflow webserver
If you want to run another executor, use the other docker-compose.yml files provided in this repository.
For LocalExecutor :
docker-compose -f docker-compose-LocalExecutor.yml up -d
For CeleryExecutor :
docker-compose -f docker-compose-CeleryExecutor.yml up -d
NB : If you want to have DAGs example loaded (default=False), you've to set the following environment variable :
LOAD_EX=n
docker run -d -p 8080:8080 -e LOAD_EX=y puckel/docker-airflow
If you want to use Ad hoc query, make sure you've configured connections: Go to Admin -> Connections and Edit "postgres_default" set this values (equivalent to values in airflow.cfg/docker-compose*.yml) :
- Host : postgres
- Schema : airflow
- Login : airflow
- Password : airflow
For encrypted connection passwords (in Local or Celery Executor), you must have the same fernet_key. By default docker-airflow generates the fernet_key at startup, you have to set an environment variable in the docker-compose (ie: docker-compose-LocalExecutor.yml) file to set the same key accross containers. To generate a fernet_key :
docker run puckel/docker-airflow python -c "from cryptography.fernet import Fernet; FERNET_KEY = Fernet.generate_key().decode(); print(FERNET_KEY)"
It's possible to set any configuration value for Airflow from environment variables, which are used over values from the airflow.cfg.
The general rule is the environment variable should be named AIRFLOW__<section>__<key>, for example AIRFLOW__CORE__SQL_ALCHEMY_CONN sets the sql_alchemy_conn config option in the [core] section.
Check out the Airflow documentation for more details
You can also define connections via environment variables by prefixing them with AIRFLOW_CONN_ - for example AIRFLOW_CONN_POSTGRES_MASTER=postgres://user:password@localhost:5432/master for a connection called "postgres_master". The value is parsed as a URI. This will work for hooks etc, but won't show up in the "Ad-hoc Query" section unless an (empty) connection is also created in the DB
Airflow allows for custom user-created plugins which are typically found in ${AIRFLOW_HOME}/plugins folder. Documentation on plugins can be found here
In order to incorporate plugins into your docker container
- Create the plugins folders
plugins/with your custom plugins. - Mount the folder as a volume by doing either of the following:
- Include the folder as a volume in command-line
-v $(pwd)/plugins/:/usr/local/airflow/plugins - Use docker-compose-LocalExecutor.yml or docker-compose-CeleryExecutor.yml which contains support for adding the plugins folder as a volume
- Include the folder as a volume in command-line
- Create a file "requirements.txt" with the desired python modules
- Mount this file as a volume
-v $(pwd)/requirements.txt:/requirements.txt(or add it as a volume in docker-compose file) - The entrypoint.sh script execute the pip install command (with --user option)
- Airflow: localhost:8080
- Flower: localhost:5555
Easy scaling using docker-compose:
docker-compose -f docker-compose-CeleryExecutor.yml scale worker=5
This can be used to scale to a multi node setup using docker swarm.
If you want to run other airflow sub-commands, such as list_dags or clear you can do so like this:
docker run --rm -ti puckel/docker-airflow airflow list_dags
or with your docker-compose set up like this:
docker-compose -f docker-compose-CeleryExecutor.yml run --rm webserver airflow list_dags
You can also use this to run a bash shell or any other command in the same environment that airflow would be run in:
docker run --rm -ti puckel/docker-airflow bash
docker run --rm -ti puckel/docker-airflow ipython
If the executor type is set to anything else than SequentialExecutor you'll need an SQL database.
Here is a list of PostgreSQL configuration variables and their default values. They're used to compute
the AIRFLOW__CORE__SQL_ALCHEMY_CONN and AIRFLOW__CELERY__RESULT_BACKEND variables when needed for you
if you don't provide them explicitly:
| Variable | Default value | Role |
|---|---|---|
POSTGRES_HOST |
postgres |
Database server host |
POSTGRES_PORT |
5432 |
Database server port |
POSTGRES_USER |
airflow |
Database user |
POSTGRES_PASSWORD |
airflow |
Database password |
POSTGRES_DB |
airflow |
Database name |
POSTGRES_EXTRAS |
empty | Extras parameters |
You can also use those variables to adapt your compose file to match an existing PostgreSQL instance managed elsewhere.
Please refer to the Airflow documentation to understand the use of extras parameters, for example in order to configure a connection that uses TLS encryption.
Here's an important thing to consider:
When specifying the connection as URI (in AIRFLOW_CONN_* variable) you should specify it following the standard syntax of DB connections, where extras are passed as parameters of the URI (note that all components of the URI should be URL-encoded).
Therefore you must provide extras parameters URL-encoded, starting with a leading ?. For example:
POSTGRES_EXTRAS="?sslmode=verify-full&sslrootcert=%2Fetc%2Fssl%2Fcerts%2Fca-certificates.crt"
If the executor type is set to CeleryExecutor you'll need a Celery broker. Here is a list of Redis configuration variables
and their default values. They're used to compute the AIRFLOW__CELERY__BROKER_URL variable for you if you don't provide
it explicitly:
| Variable | Default value | Role |
|---|---|---|
REDIS_PROTO |
redis:// |
Protocol |
REDIS_HOST |
redis |
Redis server host |
REDIS_PORT |
6379 |
Redis server port |
REDIS_PASSWORD |
empty | If Redis is password protected |
REDIS_DBNUM |
1 |
Database number |
You can also use those variables to adapt your compose file to match an existing Redis instance managed elsewhere.
Fork, improve and PR.