If you are using the default entrypoint of the production image, there are a few actions that are automatically performed when the container starts. In some cases, you can pass environment variables to the image to trigger some of that behaviour.

The variables that control the “execution” behaviour start with _AIRFLOW to distinguish them from the variables used to build the image starting with AIRFLOW.

Allowing arbitrary user to run the container

Airflow image is Open-Shift compatible, which means that you can start it with random user ID and the group id 0 (root). If you want to run the image with user different than Airflow, you MUST set GID of the user to 0. In case you try to use different group, the entrypoint exits with error.

OpenShift randomly assigns UID when it starts the container, but you can utilise this flexible UID also in case of running the image manually. This might be useful for example in case you want to mount dag and logs folders from host system on Linux, in which case the UID should be set the same ID as your host user.

This can be achieved in various ways - you can change USER when you extend or customize the image or you can dynamically pass the user to docker run command, by adding --user flag in one of those formats (See Docker Run reference for details):

[ user | user:group | uid | uid:gid | user:gid | uid:group ]

In case of Docker Compose environment it can be changed via user: entry in the docker-compose.yaml. See Docker compose reference for details. In our Quickstart Guide using Docker-Compose, the UID can be passed via the AIRFLOW_UID variable as described in Initializing docker compose environment.

The user can be any UID. In case UID is different than the default airflow (UID=50000), the user will be automatically created when entering the container.

In order to accommodate a number of external libraries and projects, Airflow will automatically create such an arbitrary user in (/etc/passwd) and make its home directory point to /home/airflow. Many of 3rd-party libraries and packages require home directory of the user to be present, because they need to write some cache information there, so such a dynamic creation of a user is necessary.

Such arbitrary user has to be able to write to certain directories that needs write access, and since it is not advised to allow write access to “other” for security reasons, the OpenShift guidelines introduced the concept of making all such folders have the 0 (root) group id (GID). All the directories that need write access in the Airflow production image have GID set to 0 (and they are writable for the group). We are following that concept and all the directories that need write access follow that.

The GID=0 is set as default for the airflow user, so any directories it creates have GID set to 0 by default. The entrypoint sets umask to be 0002 - this means that any directories created by the user have also “group write” access for group 0 - they will be writable by other users with root group. Also whenever any “arbitrary” user creates a folder (for example in a mounted volume), that folder will have a “group write” access and GID=0, so that execution with another, arbitrary user will still continue to work, even if such directory is mounted by another arbitrary user later.

The umask setting, however, only works for runtime of the container - it is not used during building of the image. If you would like to extend the image and add your own packages, you should remember to add umask 0002 in front of your docker command - this way the directories created by any installation that need group access will also be writable for the group. This can be done for example this way:

RUN umask 0002; \
    do_something; \

You can read more about it in the “Support arbitrary user ids” chapter in the Openshift best practices.

Waits for Airflow DB connection

The entrypoint is waiting for a connection to the database independent of the database engine. This allows us to increase the stability of the environment.

Waiting for connection involves executing airflow db check command, which means that a select 1 as is_alive; statement is executed. Then it loops until the command will be successful. It tries CONNECTION_CHECK_MAX_COUNT times and sleeps CONNECTION_CHECK_SLEEP_TIME between checks To disable check, set CONNECTION_CHECK_MAX_COUNT=0.

Waits for Celery broker connection

In case CeleryExecutor is used, and one of the scheduler, celery commands are used the entrypoint will wait until the Celery broker DB connection is available.

The script detects backend type depending on the URL schema and assigns default port numbers if not specified in the URL. Then it loops until connection to the host/port specified can be established It tries CONNECTION_CHECK_MAX_COUNT times and sleeps CONNECTION_CHECK_SLEEP_TIME between checks. To disable check, set CONNECTION_CHECK_MAX_COUNT=0.

Supported schemes:

  • amqp(s):// (rabbitmq) - default port 5672

  • redis:// - default port 6379

  • postgres:// - default port 5432

  • mysql:// - default port 3306

Waiting for connection involves checking if a matching port is open. The host information is derived from the Airflow configuration.

Executing commands

If first argument equals to “bash” - you are dropped to a bash shell or you can executes bash command if you specify extra arguments. For example:

docker run -it apache/airflow:2.9.1-python3.8 bash -c "ls -la"
total 16
drwxr-xr-x 4 airflow root 4096 Jun  5 18:12 .
drwxr-xr-x 1 root    root 4096 Jun  5 18:12 ..
drwxr-xr-x 2 airflow root 4096 Jun  5 18:12 dags
drwxr-xr-x 2 airflow root 4096 Jun  5 18:12 logs

If first argument is equal to python - you are dropped in python shell or python commands are executed if you pass extra parameters. For example:

> docker run -it apache/airflow:2.9.1-python3.8 python -c "print('test')"

If first argument equals to “airflow” - the rest of the arguments is treated as an airflow command to execute. Example:

docker run -it apache/airflow:2.9.1-python3.8 airflow webserver

If there are any other arguments - they are simply passed to the “airflow” command

> docker run -it apache/airflow:2.9.1-python3.8 help
  usage: airflow [-h] GROUP_OR_COMMAND ...

  positional arguments:

        celery         Celery components
        config         View configuration
        connections    Manage connections
        dags           Manage DAGs
        db             Database operations
        jobs           Manage jobs
        kubernetes     Tools to help run the KubernetesExecutor
        pools          Manage pools
        providers      Display providers
        roles          Manage roles
        tasks          Manage tasks
        users          Manage users
        variables      Manage variables

        cheat-sheet    Display cheat sheet
        info           Show information about current Airflow and environment
        kerberos       Start a Kerberos ticket renewer
        plugins        Dump information about loaded plugins
                       Rotate encrypted connection credentials and variables
        scheduler      Start a scheduler instance
        sync-perm      Update permissions for existing roles and optionally DAGs
        version        Show the version
        webserver      Start a Airflow webserver instance

  optional arguments:
    -h, --help         show this help message and exit

Execute custom code before the Airflow entrypoint

If you want to execute some custom code before Airflow’s entrypoint you can by using a custom script and calling Airflow’s entrypoint as the last exec instruction in your custom one. However, you have to remember to use dumb-init in the same way as it is used with Airflow’s entrypoint, otherwise you might have problems with proper signal propagation (See the next chapter).

FROM airflow:2.9.0.dev0
COPY my_entrypoint.sh /
ENTRYPOINT ["/usr/bin/dumb-init", "--", "/my_entrypoint.sh"]

Your entrypoint might for example modify or add variables on the fly. For example the below entrypoint sets max count of DB checks from the first parameter passed as parameter of the image execution (A bit useless example but should give the reader an example of how you could use it).

exec /entrypoint "${@}"

Make sure Airflow’s entrypoint is run with exec /entrypoint "${@}" as the last command in your custom entrypoint. This way signals will be properly propagated and arguments will be passed to the entrypoint as usual (you can use shift as above if you need to pass some extra arguments. Note that passing secret values this way or storing secrets inside the image is a bad idea from security point of view - as both image and parameters to run the image with are accessible to anyone who has access to logs of your Kubernetes or image registry.

Also be aware that code executed before Airflow’s entrypoint should not create any files or directories inside the container and everything might not work the same way when it is executed. Before Airflow entrypoint is executed, the following functionalities are not available:

  • umask is not set properly to allow group write access

  • user is not yet created in /etc/passwd if an arbitrary user is used to run the image

  • the database and brokers might not be available yet

Adding custom image behaviour

The Airflow image executes a lot of steps in the entrypoint, and sets the right environment, but you might want to run additional code after the entrypoint creates the user, sets the umask, sets variables and checks that database is running.

Rather than running regular commands - scheduler, webserver you can run custom script that you can embed into the image. You can even execute the usual components of airflow - scheduler, webserver in your custom script when you finish your custom setup. Similarly to custom entrypoint, it can be added to the image by extending it.

FROM airflow:2.9.0.dev0
COPY my_after_entrypoint_script.sh /

Build your image and then you can run this script by running the command:

docker build . --pull --tag my-image:0.0.1
docker run -it my-image:0.0.1 bash -c "/my_after_entrypoint_script.sh"

Signal propagation

Airflow uses dumb-init to run as “init” in the entrypoint. This is in order to propagate signals and reap child processes properly. This means that the process that you run does not have to install signal handlers to work properly and be killed when the container is gracefully terminated. The behaviour of signal propagation is configured by DUMB_INIT_SETSID variable which is set to 1 by default - meaning that the signals will be propagated to the whole process group, but you can set it to 0 to enable single-child behaviour of dumb-init which only propagates the signals to only single child process.

The table below summarizes DUMB_INIT_SETSID possible values and their use cases.

Variable value

Use case

1 (default)

Propagates signals to all processes in the process group of the main process running in the container.

If you run your processes via ["bash", "-c"] command and bash spawn new processes without exec, this will help to terminate your container gracefully as all processes will receive the signal.


Propagates signals to the main process only.

This is useful if your main process handles signals gracefully. A good example is warm shutdown of Celery workers. The dumb-init in this case will only propagate the signals to the main process, but not to the processes that are spawned in the same process group as the main one. For example in case of Celery, the main process will put the worker in “offline” mode, and will wait until all running tasks complete, and only then it will terminate all processes.

For Airflow’s Celery worker, you should set the variable to 0 and either use ["celery", "worker"] command. If you are running it through ["bash", "-c"] command, you need to start the worker via exec airflow celery worker as the last command executed.

Additional quick test options

The options below are mostly used for quick testing the image - for example with quick-start docker-compose or when you want to perform a local test with new packages added. They are not supposed to be run in the production environment as they add additional overhead for execution of additional commands. Those options in production should be realized either as maintenance operations on the database or should be embedded in the custom image used (when you want to add new packages).

Upgrading Airflow DB

If you set _AIRFLOW_DB_MIGRATE variable to a non-empty value, the entrypoint will run the airflow db migrate command right after verifying the connection. You can also use this when you are running airflow with internal SQLite database (default) to upgrade the db and create admin users at entrypoint, so that you can start the webserver immediately. Note - using SQLite is intended only for testing purpose, never use SQLite in production as it has severe limitations when it comes to concurrency.

Creating admin user

The entrypoint can also create webserver user automatically when you enter it. you need to set _AIRFLOW_WWW_USER_CREATE to a non-empty value in order to do that. This is not intended for production, it is only useful if you would like to run a quick test with the production image. You need to pass at least password to create such user via _AIRFLOW_WWW_USER_PASSWORD or _AIRFLOW_WWW_USER_PASSWORD_CMD similarly like for other *_CMD variables, the content of the *_CMD will be evaluated as shell command and its output will be set as password.

User creation will fail if none of the PASSWORD variables are set - there is no default for password for security reasons.



Environment variable


















In case the password is specified, the user will be attempted to be created, but the entrypoint will not fail if the attempt fails (this accounts for the case that the user is already created).

You can, for example start the webserver in the production image with initializing the internal SQLite database and creating an admin/admin Admin user with the following command:

docker run -it -p 8080:8080 \
  --env "_AIRFLOW_DB_MIGRATE=true" \
  --env "_AIRFLOW_WWW_USER_CREATE=true" \
  --env "_AIRFLOW_WWW_USER_PASSWORD=admin" \
    apache/airflow:2.9.1-python3.8 webserver
docker run -it -p 8080:8080 \
  --env "_AIRFLOW_DB_MIGRATE=true" \
  --env "_AIRFLOW_WWW_USER_CREATE=true" \
  --env "_AIRFLOW_WWW_USER_PASSWORD_CMD=echo admin" \
    apache/airflow:2.9.1-python3.8 webserver

The commands above perform initialization of the SQLite database, create admin user with admin password and Admin role. They also forward local port 8080 to the webserver port and finally start the webserver.

Installing additional requirements


Installing requirements this way is a very convenient method of running Airflow, very useful for testing and debugging. However, do not be tricked by its convenience. You should never, ever use it in production environment. We have deliberately chose to make it a development/test dependency and we print a warning, whenever it is used. There is an inherent security-related issue with using this method in production. Installing the requirements this way can happen at literally any time - when your containers get restarted, when your machines in K8S cluster get restarted. In a K8S Cluster those events can happen literally any time. This opens you up to a serious vulnerability where your production environment might be brought down by a single dependency being removed from PyPI - or even dependency of your dependency. This means that you put your production service availability in hands of 3rd-party developers. At any time, any moment including weekends and holidays those 3rd party developers might bring your production Airflow instance down, without you even knowing it. This is a serious vulnerability that is similar to the infamous leftpad problem. You can fully protect against this case by building your own, immutable custom image, where the dependencies are baked in. You have been warned.

Installing additional requirements can be done by specifying _PIP_ADDITIONAL_REQUIREMENTS variable. The variable should contain a list of requirements that should be installed additionally when entering the containers. Note that this option slows down starting of Airflow as every time any container starts it must install new packages and it opens up huge potential security vulnerability when used in production (see below). Therefore this option should only be used for testing. When testing is finished, you should create your custom image with dependencies baked in.


docker run -it -p 8080:8080 \
  --env "_PIP_ADDITIONAL_REQUIREMENTS=lxml==4.6.3 charset-normalizer==1.4.1" \
  --env "_AIRFLOW_DB_MIGRATE=true" \
  --env "_AIRFLOW_WWW_USER_CREATE=true" \
  --env "_AIRFLOW_WWW_USER_PASSWORD_CMD=echo admin" \
    apache/airflow:2.9.1-python3.8 webserver

This method is only available starting from Docker image of Airflow 2.1.1 and above.

Was this entry helpful?