Running Airflow in Docker¶

This quick-start guide will allow you to quickly get Airflow up and running with the CeleryExecutor in Docker.

Caution

This procedure can be useful for learning and exploration. However, adapting it for use in real-world situations can be complicated. Making changes to this procedure will require specialized expertise in Docker & Docker Compose, and the Airflow community may not be able to help you.

For that reason, we recommend using Kubernetes with the Official Airflow Community Helm Chart when you are ready to run Airflow in production.

Before you begin¶

This procedure assumes familiarity with Docker and Docker Compose. If you haven’t worked with these tools before, you should take a moment to run through the Docker Quick Start (especially the section on Docker Compose) so you are familiar with how they work.

Follow these steps to install the necessary tools, if you have not already done so.

Install Docker Community Edition (CE) on your workstation. Depending on your OS, you may need to configure Docker to use at least 4.00 GB of memory for the Airflow containers to run properly. Please refer to the Resources section in the Docker for Windows or Docker for Mac documentation for more information.
Install Docker Compose v2.14.0 or newer on your workstation.

Older versions of docker-compose do not support all the features required by the Airflow docker-compose.yaml file, so double check that your version meets the minimum version requirements.

Tip

The default amount of memory available for Docker on macOS is often not enough to get Airflow up and running. If enough memory is not allocated, it might lead to the webserver continuously restarting. You should allocate at least 4GB memory for the Docker Engine (ideally 8GB).

You can check if you have enough memory by running this command:

docker run --rm "debian:bullseye-slim" bash -c 'numfmt --to iec $(echo $(($(getconf _PHYS_PAGES) * $(getconf PAGE_SIZE))))'

Warning

Some operating systems (Fedora, ArchLinux, RHEL, Rocky) have recently introduced Kernel changes that result in Airflow in Docker Compose consuming 100% memory when run inside the community Docker implementation maintained by the OS teams.

This is an issue with backwards-incompatible containerd configuration that some of Airflow dependencies have problems with and is tracked in a few issues:

There is no solution yet from the containerd team, but seems that installing Docker Desktop on Linux solves the problem as stated in This comment and allows to run Breeze with no problems.

Fetching `docker-compose.yaml`¶

To deploy Airflow on Docker Compose, you should fetch docker-compose.yaml.

curl -LfO 'https://airflow.apache.org/docs/apache-airflow/2.9.1/docker-compose.yaml'

Important

From July 2023 Compose V1 stopped receiving updates. We strongly advise upgrading to a newer version of Docker Compose, supplied docker-compose.yaml may not function accurately within Compose V1.

This file contains several service definitions:

airflow-scheduler - The scheduler monitors all tasks and DAGs, then triggers the task instances once their dependencies are complete.
airflow-webserver - The webserver is available at http://localhost:8080.
airflow-worker - The worker that executes the tasks given by the scheduler.
airflow-triggerer - The triggerer runs an event loop for deferrable tasks.
airflow-init - The initialization service.
postgres - The database.
redis - The redis - broker that forwards messages from scheduler to worker.

Optionally, you can enable flower by adding --profile flower option, e.g. docker compose --profile flower up, or by explicitly specifying it on the command line e.g. docker compose up flower.

flower - The flower app for monitoring the environment. It is available at http://localhost:5555.

All these services allow you to run Airflow with CeleryExecutor. For more information, see Architecture Overview.

Some directories in the container are mounted, which means that their contents are synchronized between your computer and the container.

./dags - you can put your DAG files here.
./logs - contains logs from task execution and scheduler.
./config - you can add custom log parser or add airflow_local_settings.py to configure cluster policy.
./plugins - you can put your custom plugins here.

This file uses the latest Airflow image (apache/airflow). If you need to install a new Python library or system library, you can build your image.

Initializing Environment¶

Before starting Airflow for the first time, you need to prepare your environment, i.e. create the necessary files, directories and initialize the database.

Setting the right Airflow user¶

On Linux, the quick-start needs to know your host user id and needs to have group id set to 0. Otherwise the files created in dags, logs and plugins will be created with root user ownership. You have to make sure to configure them for the docker-compose:

mkdir -p ./dags ./logs ./plugins ./config
echo -e "AIRFLOW_UID=$(id -u)" > .env

See Docker Compose environment variables

For other operating systems, you may get a warning that AIRFLOW_UID is not set, but you can safely ignore it. You can also manually create an .env file in the same folder as docker-compose.yaml with this content to get rid of the warning:

AIRFLOW_UID=50000

Initialize the database¶

On all operating systems, you need to run database migrations and create the first user account. To do this, run.

docker compose up airflow-init

After initialization is complete, you should see a message like this:

airflow-init_1       | Upgrades done
airflow-init_1       | Admin user airflow created
airflow-init_1       | 2.9.1
start_airflow-init_1 exited with code 0

The account created has the login airflow and the password airflow.

Cleaning-up the environment¶

The docker-compose environment we have prepared is a “quick-start” one. It was not designed to be used in production and it has a number of caveats - one of them being that the best way to recover from any problem is to clean it up and restart from scratch.

The best way to do this is to:

Run docker compose down --volumes --remove-orphans command in the directory you downloaded the docker-compose.yaml file
Remove the entire directory where you downloaded the docker-compose.yaml file rm -rf '<DIRECTORY>'
Run through this guide from the very beginning, starting by re-downloading the docker-compose.yaml file

Running Airflow¶

Now you can start all services:

docker compose up

Note

docker-compose is old syntax. Please check Stackoverflow.

In a second terminal you can check the condition of the containers and make sure that no containers are in an unhealthy condition:

$ docker ps
CONTAINER ID   IMAGE                  COMMAND                  CREATED          STATUS                    PORTS                              NAMES
247ebe6cf87a   apache/airflow:2.9.1   "/usr/bin/dumb-init …"   3 minutes ago    Up 3 minutes (healthy)    8080/tcp                           compose_airflow-worker_1
ed9b09fc84b1   apache/airflow:2.9.1   "/usr/bin/dumb-init …"   3 minutes ago    Up 3 minutes (healthy)    8080/tcp                           compose_airflow-scheduler_1
7cb1fb603a98   apache/airflow:2.9.1   "/usr/bin/dumb-init …"   3 minutes ago    Up 3 minutes (healthy)    0.0.0.0:8080->8080/tcp             compose_airflow-webserver_1
74f3bbe506eb   postgres:13            "docker-entrypoint.s…"   18 minutes ago   Up 17 minutes (healthy)   5432/tcp                           compose_postgres_1
0bd6576d23cb   redis:latest           "docker-entrypoint.s…"   10 hours ago     Up 17 minutes (healthy)   0.0.0.0:6379->6379/tcp             compose_redis_1

Accessing the environment¶

After starting Airflow, you can interact with it in 3 ways:

by running CLI commands.
via a browser using the web interface.
using the REST API.

Running the CLI commands¶

You can also run CLI commands, but you have to do it in one of the defined airflow-* services. For example, to run airflow info, run the following command:

docker compose run airflow-worker airflow info

If you have Linux or Mac OS, you can make your work easier and download a optional wrapper scripts that will allow you to run commands with a simpler command.

curl -LfO 'https://airflow.apache.org/docs/apache-airflow/2.9.1/airflow.sh'
chmod +x airflow.sh

Now you can run commands easier.

./airflow.sh info

You can also use bash as parameter to enter interactive bash shell in the container or python to enter python container.

./airflow.sh bash

./airflow.sh python

Accessing the web interface¶

Once the cluster has started up, you can log in to the web interface and begin experimenting with DAGs.

The webserver is available at: http://localhost:8080. The default account has the login airflow and the password airflow.

Sending requests to the REST API¶

Basic username password authentication is currently supported for the REST API, which means you can use common tools to send requests to the API.

The webserver is available at: http://localhost:8080. The default account has the login airflow and the password airflow.

Here is a sample curl command, which sends a request to retrieve a pool list:

ENDPOINT_URL="http://localhost:8080/"
curl -X GET  \
    --user "airflow:airflow" \
    "${ENDPOINT_URL}/api/v1/pools"

Cleaning up¶

To stop and delete containers, delete volumes with database data and download images, run:

docker compose down --volumes --rmi all

Using custom images¶

When you want to run Airflow locally, you might want to use an extended image, containing some additional dependencies - for example you might add new python packages, or upgrade airflow providers to a later version. This can be done very easily by specifying build: . in your docker-compose.yaml and placing a custom Dockerfile alongside your docker-compose.yaml. Then you can use docker compose build command to build your image (you need to do it only once). You can also add the --build flag to your docker compose commands to rebuild the images on-the-fly when you run other docker compose commands.

Examples of how you can extend the image with custom providers, python packages, apt packages and more can be found in Building the image.

Note

Creating custom images means that you need to maintain also a level of automation as you need to re-create the images when either the packages you want to install or Airflow is upgraded. Please do not forget about keeping these scripts. Also keep in mind, that in cases when you run pure Python tasks, you can use the Python Virtualenv functions which will dynamically source and install python dependencies during runtime. With Airflow 2.8.0 Virtualenvs can also be cached.

Special case - adding dependencies via requirements.txt file¶

Usual case for custom images, is when you want to add a set of requirements to it - usually stored in requirements.txt file. For development, you might be tempted to add it dynamically when you are starting the original airflow image, but this has a number of side effects (for example your containers will start much slower - each additional dependency will further delay your containers start up time). Also it is completely unnecessary, because docker compose has the development workflow built-in. You can - following the previous chapter, automatically build and use your custom image when you iterate with docker compose locally. Specifically when you want to add your own requirement file, you should do those steps:

Comment out the image: ... line and remove comment from the build: . line in the docker-compose.yaml file. The relevant part of the docker-compose file of yours should look similar to (use correct image tag):

#image: ${AIRFLOW_IMAGE_NAME:-apache/airflow:2.9.1}
build: .

Create Dockerfile in the same folder your docker-compose.yaml file is with content similar to:

FROM apache/airflow:2.9.1
ADD requirements.txt .
RUN pip install apache-airflow==${AIRFLOW_VERSION} -r requirements.txt

It is the best practice to install apache-airflow in the same version as the one that comes from the original image. This way you can be sure that pip will not try to downgrade or upgrade apache airflow while installing other requirements, which might happen in case you try to add a dependency that conflicts with the version of apache-airflow that you are using.

Place requirements.txt file in the same directory.

Run docker compose build to build the image, or add --build flag to docker compose up or docker compose run commands to build the image automatically as needed.

Special case - Adding a custom config file¶

If you have a custom config file and wish to use it in your Airflow instance, you need to perform the following steps:

Remove comment from the AIRFLOW_CONFIG: '/opt/airflow/config/airflow.cfg' line in the docker-compose.yaml file.
Place your custom airflow.cfg file in the local config folder.
If your config file has a different name than airflow.cfg, adjust the filename in AIRFLOW_CONFIG: '/opt/airflow/config/airflow.cfg'

Networking¶

In general, if you want to use Airflow locally, your DAGs may try to connect to servers which are running on the host. In order to achieve that, an extra configuration must be added in docker-compose.yaml. For example, on Linux the configuration must be in the section services: airflow-worker adding extra_hosts: - "host.docker.internal:host-gateway"; and use host.docker.internal instead of localhost. This configuration vary in different platforms. Please check the Docker documentation for Windows and Mac for further information.

Debug Airflow inside docker container using PyCharm¶

Prerequisites: Create a project in PyCharm and download the (docker-compose.yaml).

Steps:

Modify docker-compose.yaml

Add the following section under the services section:

airflow-python:
<<: *airflow-common
profiles:
    - debug
environment:
    <<: *airflow-common-env
user: "50000:0"
entrypoint: ["bash"]

Note

This code snippet creates a new service named “airflow-python” specifically for PyCharm’s Python interpreter. On a Linux system, if you have executed the command echo -e "AIRFLOW_UID=$(id -u)" > .env, you need to set user: "50000:0" in airflow-python service to avoid PyCharm’s Unresolved reference 'airflow' error.

Configure PyCharm Interpreter
- Open PyCharm and navigate to Settings (or Preferences on macOS) > Project: <Your Project Name> > Python Interpreter.
- Click the “Add Interpreter” button and choose “On Docker Compose”.
- In the Configuration file field, select your docker-compose.yaml file.
- In the Service field, choose the newly added airflow-python service.
- Click “Next” and follow the prompts to complete the configuration.

Configuring the container's Python interpreter in PyCharm, step diagram

Building the interpreter index might take some time. Once configured, you can debug your Airflow code within the container environment, mimicking your local setup.

FAQ: Frequently asked questions¶

`ModuleNotFoundError: No module named 'XYZ'`¶

The Docker Compose file uses the latest Airflow image (apache/airflow). If you need to install a new Python library or system library, you can customize and extend it.

What’s Next?¶

From this point, you can head to the Tutorials section for further examples or the How-to Guides section if you’re ready to get your hands dirty.

Environment variables supported by Docker Compose¶

Do not confuse the variable names here with the build arguments set when image is built. The AIRFLOW_UID build arg defaults to 50000 when the image is built, so it is “baked” into the image. On the other hand, the environment variables below can be set when the container is running, using - for example - result of id -u command, which allows to use the dynamic host runtime user id which is unknown at the time of building the image.

Variable	Description	Default
`AIRFLOW_IMAGE_NAME`	Airflow Image to use.	apache/airflow:2.9.1
`AIRFLOW_UID`	UID of the user to run Airflow containers as. Override if you want to use non-default Airflow UID (for example when you map folders from host, it should be set to result of `id -u` call. When it is changed, a user with the UID is created with `default` name inside the container and home of the use is set to `/airflow/home/` in order to share Python libraries installed there. This is in order to achieve the OpenShift compatibility. See more in the Arbitrary Docker User	`50000`

Note

Before Airflow 2.2, the Docker Compose also had AIRFLOW_GID parameter, but it did not provide any additional functionality - only added confusion - so it has been removed.

Those additional variables are useful in case you are trying out/testing Airflow installation via Docker Compose. They are not intended to be used in production, but they make the environment faster to bootstrap for first time users with the most common customizations.

Variable	Description	Default
`_AIRFLOW_WWW_USER_USERNAME`	Username for the administrator UI account. If this value is specified, admin UI user gets created automatically. This is only useful when you want to run Airflow for a test-drive and want to start a container with embedded development database.	airflow
`_AIRFLOW_WWW_USER_PASSWORD`	Password for the administrator UI account. Only used when `_AIRFLOW_WWW_USER_USERNAME` set.	airflow
`_PIP_ADDITIONAL_REQUIREMENTS`	If not empty, airflow containers will attempt to install requirements specified in the variable. example: `lxml==4.6.3 charset-normalizer==1.4.1`. Available in Airflow image 2.1.1 and above.

Running Airflow in Docker¶

Before you begin¶

Fetching docker-compose.yaml¶

Initializing Environment¶

Setting the right Airflow user¶

Initialize the database¶

Cleaning-up the environment¶

Running Airflow¶

Accessing the environment¶

Running the CLI commands¶

Accessing the web interface¶

Sending requests to the REST API¶

Cleaning up¶

Using custom images¶

Special case - adding dependencies via requirements.txt file¶

Special case - Adding a custom config file¶

Networking¶

Debug Airflow inside docker container using PyCharm¶

FAQ: Frequently asked questions¶

ModuleNotFoundError: No module named 'XYZ'¶

What’s Next?¶

Environment variables supported by Docker Compose¶

Fetching `docker-compose.yaml`¶

`ModuleNotFoundError: No module named 'XYZ'`¶