Using a startup script with Amazon MWAA - Amazon Managed Workflows for Apache Airflow
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Using a startup script with Amazon MWAA

A startup script is a shell (.sh) script that you host in your environment's Amazon S3 bucket similar to your DAGs, requirements, and plugins. Amazon MWAA runs this script during startup on every individual Apache Airflow component (worker, scheduler, and web server) before installing requirements and initializing the Apache Airflow process. Use a startup script to do the following:

  • Install runtimes – Install Linux runtimes required by your workflows and connections.

  • Configure environment variables – Set environment variables for each Apache Airflow component. Overwrite common variables such as PATH, PYTHONPATH, and LD_LIBRARY_PATH.

  • Manage keys and tokens – Pass access tokens for custom repositories to requirements.txt and configure security keys.

The following topics describe how to configure a startup script to install Linux runtimes, set environment variables, and troubleshoot related issues using CloudWatch Logs.

Configure a startup script

To use a startup script with your existing Amazon MWAA environment, upload a .sh file to your environment's Amazon S3 bucket. Then, to associate the script with the environment, specify the following in your environment details:

  • The Amazon S3 URL path to the script – The relative path to the script hosted in your bucket, for example, s3://mwaa-environment/

  • The Amazon S3 version ID of the script – The version of the startup shell script in your Amazon S3 bucket. You must specify the version ID that Amazon S3 assigns to the file every time you update the script. Version IDs are Unicode, UTF-8 encoded, URL-ready, opaque strings that are no more than 1,024 bytes long, for example, 3sL4kqtJlcpXroDTDmJ+rmSpXd3dIbrHY+MTRCxf3vjVBH40Nr8X8gdRQBpUMLUo.

To complete the steps in this section, use the following sample script. The script outputs the value assigned to MWAA_AIRFLOW_COMPONENT. This environment variable identifies each Apache Airflow component that the script runs on.

Copy the code and save it locally as

#!/bin/sh ​ echo "Printing Apache Airflow component" echo $MWAA_AIRFLOW_COMPONENT

Next, upload the script to your Amazon S3 bucket.

Amazon Web Services Management Console
To upload a shell script (console)
  1. Sign in to the Amazon Web Services Management Console and open the Amazon S3 console at

  2. From the Buckets list, choose the name of the bucket associated with your environment.

  3. On the Objects tab, choose Upload.

  4. On the Upload page, drag and drop the shell script you created.

  5. Choose Upload.

The script appears in the list of Objects. Amazon S3 creates a new version ID for the file. If you update the script and upload it again using the same file name, a new version ID is assigned to the file.

Amazon CLI
To create and upload a shell script (CLI)
  1. Open a new command prompt, and run the Amazon S3 ls command to list and identify the bucket associated with your environment.

    $ aws s3 ls
  2. Navigate to the folder where you saved the shell script. Use cp in a new prompt window to upload the script to your bucket. Replace your-s3-bucket with your information.

    $ aws s3 cp s3://your-s3-bucket/

    If successful, Amazon S3 outputs the URL path to the object:

    upload: ./ to s3://your-s3-bucket/
  3. Use the following command to retrieve the latest version ID for the script.

    $ aws s3api list-object-versions --bucket your-s3-bucket --prefix startup --query 'Versions[?IsLatest].[VersionId]' --output text

You specify this version ID when you associate the script with an environment.

Now, associate the script with your environment.

Amazon Web Services Management Console
To associate the script with an environment (console)
  1. Open the Environments page on the Amazon MWAA console.

  2. Select the row for the environment you want to update, then choose Edit.

  3. On the Specify details page, for Startup script file - optional, enter the Amazon S3 URL for the script, for example: s3://your-mwaa-bucket/startup-sh..

  4. Choose the latest version from the drop down list, or Browse S3 to find the script.

  5. Choose Next, then proceed to the Review and save page.

  6. Review changes, then choose Save.

Environment updates can take between 10 to 30 minutes. Amazon MWAA runs the startup script as each component in your environment restarts.

Amazon CLI
To associate the script with an environment (CLI)
  • Open a command prompt and use update-environment to specify the Amazon S3 URL and version ID for the script.

    $ aws mwaa update-environment \ --name your-mwaa-environment \ --startup-script-s3-path \ --startup-script-s3-object-version BbdVMmBRjtestta1EsVnbybZp1Wqh1J4

    If successful, Amazon MWAA returns the Amazon Resource Name (ARN) for the environment:


Environment update can take between 10 to 30 minutes. Amazon MWAA runs the startup script as each component in your environment restarts.

Finally, retrieve log events to verify that the script is working as expected. When you activate logging for an each Apache Airflow component, Amazon MWAA creates a new log group and log stream. For more information, see Apache Airflow log types.

Amazon Web Services Management Console
To check the Apache Airflow log stream (console)
  1. Open the Environments page on the Amazon MWAA console.

  2. Choose your environment.

  3. In the Monitoring pane, choose the log group for which you want to view logs, for example, Airflow scheduler log group .

  4. In the CloudWatch console, from the Log streams list, choose a stream with the following prefix: startup_script_exection_ip.

  5. On the Log events pane, you will see the output of the command printing the value for MWAA_AIRFLOW_COMPONENT. For example, for scheduler logs, you will the following:

    Printing Apache Airflow component
    Finished running startup script. Execution time: 0.004s.
    Running verification
    Verification completed

You can repeat the previous steps to view worker and web server logs.

Install Linux runtimes using a startup script

Use a startup script to update the operating system of an Apache Airflow component, and install additional runtime libraries to use with your workflows. For example, the following script runs yum update to update the operating system.

When running yum update in a startup script, you must exclude Python using --exclude=python* as shown in the example. For your environment to run, Amazon MWAA installs a specific version of Python compatible with your environment. Therefore, you can't update the environment's Python version using a startup script.

#!/bin/sh echo "Updating operating system" sudo yum update -y --exclude=python*

To install runtimes on specific Apache Airflow component, use MWAA_AIRFLOW_COMPONENT and if and fi conditional statements. This example runs a single command to install the libaio library on the scheduler and worker, but not on the web server.

  • If you have configured a private web server, you must either use the following condition or provide all installation files locally in order to avoid installation timeouts.

  • Use sudo to run operations that require administrative privileges.

#!/bin/sh if [[ "${MWAA_AIRFLOW_COMPONENT}" != "webserver" ]] then sudo yum -y install libaio fi

You can use a startup script to check the Python version.

#!/bin/sh export PYTHON_VERSION_CHECK=`python -c 'import sys; version=sys.version_info[:3]; print("{0}.{1}.{2}".format(*version))'` echo "Python version is $PYTHON_VERSION_CHECK"

Amazon MWAA does not support overriding the default Python version, as this may lead to incompatibilities with the installed Apache Airflow libraries.

Set environment variables using a startup script

Use startup scripts to set environment variables and modify Apache Airflow configurations. The following defines a new variable, ENVIRONMENT_STAGE. You can reference this variable in a DAG or in your custom modules.

#!/bin/sh export ENVIRONMENT_STAGE="development" echo "$ENVIRONMENT_STAGE"

Use startup scripts to overwrite common Apache Airflow or system variables. For example, you set LD_LIBRARY_PATH to instruct Python to look for binaries in the path you specify. This lets you provide custom binaries for your workflows using plugins:

#!/bin/sh export LD_LIBRARY_PATH=/usr/local/airflow/plugins/your-custom-binary

Reserved environment variables

Amazon MWAA reserves a set of critical environment variables. If you overwrite a reserved variable, Amazon MWAA restores it to its default. The following lists the reserved variables:

  • MWAA__AIRFLOW__COMPONENT – Used to identify the Apache Airflow component with one of the following values: scheduler, worker, or webserver.

  • AIRFLOW__WEBSERVER__SECRET_KEY – The secret key used for securely signing session cookies in the Apache Airflow web server.

  • AIRFLOW__CORE__FERNET_KEY – The key used for encryption and decryption of sensitive data stored in the metadata database, for example, connection passwords.

  • AIRFLOW_HOME – The path to the Apache Airflow home directory where configuration files and DAG files are stored locally.

  • AIRFLOW__CELERY__BROKER_URL – The URL of the message broker used for communication between the Apache Airflow scheduler and the Celery worker nodes.

  • AIRFLOW__CELERY__RESULT_BACKEND – The URL of the database used to store the results of Celery tasks.

  • AIRFLOW__CORE__EXECUTOR – The executor class that Apache Airflow should use. In Amazon MWAA this is a CeleryExecutor

  • AIRFLOW__CORE__LOAD_EXAMPLES – Used to activate, or deactivate, the loading of example DAGs.

  • AIRFLOW__METRICS__METRICS_BLOCK_LIST – Used to manage which Apache Airflow metrics are emitted and captured by Amazon MWAA in CloudWatch.

  • SQL_ALCHEMY_CONN – The connection string for the RDS for PostgreSQL database used to store Apache Airflow metadata in Amazon MWAA.

  • AIRFLOW__CORE__SQL_ALCHEMY_CONN – Used for the same purpose as SQL_ALCHEMY_CONN, but following the new Apache Airflow naming convention.

  • AIRFLOW__CELERY__DEFAULT_QUEUE – The default queue for Celery tasks in Apache Airflow.

  • AIRFLOW__OPERATORS__DEFAULT_QUEUE – The default queue for tasks using specific Apache Airflow operators.

  • AIRFLOW_VERSION – The Apache Airflow version installed in the Amazon MWAA environment.

  • AIRFLOW_CONN_AWS_DEFAULT – The default Amazon credentials used to integrate with other Amazon services in.

  • AWS_DEFAULT_REGION – Sets the default Amazon Region used with default credentials to integrate with other Amazon services.

  • AWS_REGION – If defined, this environment variable overrides the values in the environment variable AWS_DEFAULT_REGION and the profile setting region.

  • PYTHONUNBUFFERED – Used to send stdout and stderr streams to container logs.

  • AIRFLOW__METRICS__STATSD_ALLOW_LIST – Used to configure an allow list of comma-separated prefixes to send the metrics that start with the elements of the list.

  • AIRFLOW__METRICS__STATSD_ON – Activates sending metrics to StatsD.

  • AIRFLOW__METRICS__STATSD_HOST – Used to connect to the StatSD daemon.

  • AIRFLOW__METRICS__STATSD_PORT – Used to connect to the StatSD daemon.

  • AIRFLOW__METRICS__STATSD_PREFIX – Used to connect to the StatSD daemon.

  • AIRFLOW__CELERY__WORKER_AUTOSCALE – Sets the maximum and minimum concurrency.

  • AIRFLOW__CORE__DAG_CONCURRENCY – Sets the number of task instances that can run concurrently by the scheduler in one DAG.

  • AIRFLOW__CORE__MAX_ACTIVE_TASKS_PER_DAG – Sets the maximum number of active tasks per DAG.

  • AIRFLOW__CORE__PARALLELISM – Defines the maximum number of task instances that can simultaneously.

  • AIRFLOW__SCHEDULER__PARSING_PROCESSES – Sets the maximum number of processes parsed by the scheduler to schedule DAGs.

  • AIRFLOW__CELERY_BROKER_TRANSPORT_OPTIONS__VISIBILITY_TIMEOUT – Defines the number of seconds a worker waits to acknowledge the task before the message is redelivered to another worker.

  • AIRFLOW__CELERY_BROKER_TRANSPORT_OPTIONS__REGION – Sets the Amazon Region for the underlying Celery transport.

  • AIRFLOW__CELERY_BROKER_TRANSPORT_OPTIONS__PREDEFINED_QUEUES – Sets the queue for the underlying Celery transport.

  • AIRFLOW_SCHEDULER_ALLOWED_RUN_ID_PATTERN – Used to verify the validity of your input for the run_id parameter when triggering a DAG.

  • AIRFLOW__WEBSERVER__BASE_URL – The URL of the web server used to host the Apache Airflow UI.

Unreserved environment variables

You can use a startup script to overwrite unreserved environment variables. The following lists some of these common variables:

  • PATH – Specifies a list of directories where the operating system searches for executable files and scripts. When a command runs in the command line, the system checks the directories in PATH in order to find and execute the command. When you create custom operators or tasks in Apache Airflow, you might need to rely on external scripts or executables. If the directories containing these files are not in the specified in the PATH variable, the tasks fail to run when the system is unable to locate them. By adding the appropriate directories to PATH, Apache Airflow tasks can find and run the required executables.

  • PYTHONPATH – Used by the Python interpreter to determine which directories to search for imported modules and packages. It is a list of directories that you can add to the default search path. This lets the interpreter find and load Python libraries not included in the standard library, or installed in system directories. Use this variable to add your modules and custom Python packages and use them with your DAGs.

  • LD_LIBRARY_PATH – An environment variable used by the dynamic linker and loader in Linux to find and load shared libraries. It specifies a list of directories containing shared libraries, which are searched before the default system library directories. Use this variable to specify your custom binaries.

  • CLASSPATH – Used by the Java Runtime Environment (JRE) and Java Development Kit (JDK) to locate and load Java classes, libraries, and resources at runtime. It is a list of directories, JAR files, and ZIP archives that contain compiled Java code.