Installing custom plugins - 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).

Installing custom plugins

Amazon Managed Workflows for Apache Airflow supports Apache Airflow's built-in plugin manager, allowing you to use custom Apache Airflow operators, hooks, sensors, or interfaces. This page describes the steps to install Apache Airflow custom plugins on your Amazon MWAA environment using a plugins.zip file.

Prerequisites

You'll need the following before you can complete the steps on this page.

  • Permissions — Your Amazon account must have been granted access by your administrator to the AmazonMWAAFullConsoleAccess access control policy for your environment. In addition, your Amazon MWAA environment must be permitted by your execution role to access the Amazon resources used by your environment.

  • Access — If you require access to public repositories to install dependencies directly on the web server, your environment must be configured with public network web server access. For more information, see Apache Airflow access modes.

  • Amazon S3 configuration — The Amazon S3 bucket used to store your DAGs, custom plugins in plugins.zip, and Python dependencies in requirements.txt must be configured with Public Access Blocked and Versioning Enabled.

How it works

To run custom plugins on your environment, you must do three things:

  1. Create a plugins.zip file locally.

  2. Upload the local plugins.zip file to your Amazon S3 bucket.

  3. Specify the version of this file in the Plugins file field on the Amazon MWAA console.

Note

If this is the first time you're uploading a plugins.zip to your Amazon S3 bucket, you also need to specify the path to the file on the Amazon MWAA console. You only need to complete this step once.

When to use the plugins

Plugins are required only for extending the Apache Airflow user interface, as outlined in the Apache Airflow documentation. Custom operators can be placed directly in the /dags folder alongside your DAG code.

If you need to create your own integrations with external systems, place them in the /dags folder or a subfolder within it, but not in the plugins.zip folder. In Apache Airflow 2.x, plugins are primarily used for extending the UI.

Similarly, other dependencies should not be placed in plugins.zip. Instead, they can be stored in a location under the Amazon S3 /dags folder, where they will be synchronized to each Amazon MWAA container before Apache Airflow starts.

Note

Any file in the /dags folder or in plugins.zip that does not explicitly define an Apache Airflow DAG object must be listed in an .airflowignore file.

Custom plugins overview

Apache Airflow's built-in plugin manager can integrate external features to its core by simply dropping files in an $AIRFLOW_HOME/plugins folder. It allows you to use custom Apache Airflow operators, hooks, sensors, or interfaces. The following section provides an example of flat and nested directory structures in a local development environment and the resulting import statements, which determines the directory structure within a plugins.zip.

Custom plugins directory and size limits

The Apache Airflow Scheduler and the Workers look for custom plugins during startup on the Amazon-managed Fargate container for your environment at /usr/local/airflow/plugins/*.

  • Directory structure. The directory structure (at /*) is based on the contents of your plugins.zip file. For example, if your plugins.zip contains the operators directory as a top-level directory, then the directory will be extracted to /usr/local/airflow/plugins/operators on your environment.

  • Size limit. We recommend a plugins.zip file less than than 1 GB. The larger the size of a plugins.zip file, the longer the startup time on an environment. Although Amazon MWAA doesn't limit the size of a plugins.zip file explicitly, if dependencies can't be installed within ten minutes, the Fargate service will time-out and attempt to rollback the environment to a stable state.

Note

For environments using Apache Airflow v1.10.12 or Apache Airflow v2.0.2, Amazon MWAA limits outbound traffic on the Apache Airflow web server, and does not allow you to install plugins nor Python dependencies directly on the web server. Starting with Apache Airflow v2.2.2, Amazon MWAA can install plugins and dependencies directly on the web server.

Examples of custom plugins

The following section uses sample code in the Apache Airflow reference guide to show how to structure your local development environment.

Example using a flat directory structure in plugins.zip

Apache Airflow v2

The following example shows a plugins.zip file with a flat directory structure for Apache Airflow v2.

Example flat directory with PythonVirtualenvOperator plugins.zip

The following example shows the top-level tree of a plugins.zip file for the PythonVirtualenvOperator custom plugin in Creating a custom plugin for Apache Airflow PythonVirtualenvOperator.

├── virtual_python_plugin.py
Example plugins/virtual_python_plugin.py

The following example shows the PythonVirtualenvOperator custom plugin.

""" Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. """ from airflow.plugins_manager import AirflowPlugin import airflow.utils.python_virtualenv from typing import List def _generate_virtualenv_cmd(tmp_dir: str, python_bin: str, system_site_packages: bool) -> List[str]: cmd = ['python3','/usr/local/airflow/.local/lib/python3.7/site-packages/virtualenv', tmp_dir] if system_site_packages: cmd.append('--system-site-packages') if python_bin is not None: cmd.append(f'--python={python_bin}') return cmd airflow.utils.python_virtualenv._generate_virtualenv_cmd=_generate_virtualenv_cmd class VirtualPythonPlugin(AirflowPlugin): name = 'virtual_python_plugin'
Apache Airflow v1

The following example shows a plugins.zip file with a flat directory structure for Apache Airflow v1.

Example flat directory with PythonVirtualenvOperator plugins.zip

The following example shows the top-level tree of a plugins.zip file for the PythonVirtualenvOperator custom plugin in Creating a custom plugin for Apache Airflow PythonVirtualenvOperator.

├── virtual_python_plugin.py
Example plugins/virtual_python_plugin.py

The following example shows the PythonVirtualenvOperator custom plugin.

from airflow.plugins_manager import AirflowPlugin from airflow.operators.python_operator import PythonVirtualenvOperator def _generate_virtualenv_cmd(self, tmp_dir): cmd = ['python3','/usr/local/airflow/.local/lib/python3.7/site-packages/virtualenv', tmp_dir] if self.system_site_packages: cmd.append('--system-site-packages') if self.python_version is not None: cmd.append('--python=python{}'.format(self.python_version)) return cmd PythonVirtualenvOperator._generate_virtualenv_cmd=_generate_virtualenv_cmd class EnvVarPlugin(AirflowPlugin): name = 'virtual_python_plugin'

Example using a nested directory structure in plugins.zip

Apache Airflow v2

The following example shows a plugins.zip file with separate directories for hooks, operators, and a sensors directory for Apache Airflow v2.

Example plugins.zip
__init__.py my_airflow_plugin.py hooks/ |-- __init__.py |-- my_airflow_hook.py operators/ |-- __init__.py |-- my_airflow_operator.py |-- hello_operator.py sensors/ |-- __init__.py |-- my_airflow_sensor.py

The following example shows the import statements in the DAG (DAGs folder) that uses the custom plugins.

Example dags/your_dag.py
from airflow import DAG from datetime import datetime, timedelta from operators.my_airflow_operator import MyOperator from sensors.my_airflow_sensor import MySensor from operators.hello_operator import HelloOperator default_args = { 'owner': 'airflow', 'depends_on_past': False, 'start_date': datetime(2018, 1, 1), 'email_on_failure': False, 'email_on_retry': False, 'retries': 1, 'retry_delay': timedelta(minutes=5), } with DAG('customdag', max_active_runs=3, schedule_interval='@once', default_args=default_args) as dag: sens = MySensor( task_id='taskA' ) op = MyOperator( task_id='taskB', my_field='some text' ) hello_task = HelloOperator(task_id='sample-task', name='foo_bar') sens >> op >> hello_task
Example plugins/my_airflow_plugin.py
from airflow.plugins_manager import AirflowPlugin from hooks.my_airflow_hook import * from operators.my_airflow_operator import * class PluginName(AirflowPlugin): name = 'my_airflow_plugin' hooks = [MyHook] operators = [MyOperator] sensors = [MySensor]

The following examples show each of the import statements needed in the custom plugin files.

Example hooks/my_airflow_hook.py
from airflow.hooks.base import BaseHook class MyHook(BaseHook): def my_method(self): print("Hello World")
Example sensors/my_airflow_sensor.py
from airflow.sensors.base import BaseSensorOperator from airflow.utils.decorators import apply_defaults class MySensor(BaseSensorOperator): @apply_defaults def __init__(self, *args, **kwargs): super(MySensor, self).__init__(*args, **kwargs) def poke(self, context): return True
Example operators/my_airflow_operator.py
from airflow.operators.bash import BaseOperator from airflow.utils.decorators import apply_defaults from hooks.my_airflow_hook import MyHook class MyOperator(BaseOperator): @apply_defaults def __init__(self, my_field, *args, **kwargs): super(MyOperator, self).__init__(*args, **kwargs) self.my_field = my_field def execute(self, context): hook = MyHook('my_conn') hook.my_method()
Example operators/hello_operator.py
from airflow.models.baseoperator import BaseOperator from airflow.utils.decorators import apply_defaults class HelloOperator(BaseOperator): @apply_defaults def __init__( self, name: str, **kwargs) -> None: super().__init__(**kwargs) self.name = name def execute(self, context): message = "Hello {}".format(self.name) print(message) return message

Follow the steps in Testing custom plugins using the Amazon MWAA CLI utility, and then Creating a plugins.zip file to zip the contents within your plugins directory. For example, cd plugins.

Apache Airflow v1

The following example shows a plugins.zip file with separate directories for hooks, operators, and a sensors directory for Apache Airflow v1.10.12.

Example plugins.zip
__init__.py my_airflow_plugin.py hooks/ |-- __init__.py |-- my_airflow_hook.py operators/ |-- __init__.py |-- my_airflow_operator.py |-- hello_operator.py sensors/ |-- __init__.py |-- my_airflow_sensor.py

The following example shows the import statements in the DAG (DAGs folder) that uses the custom plugins.

Example dags/your_dag.py
from airflow import DAG from datetime import datetime, timedelta from operators.my_operator import MyOperator from sensors.my_sensor import MySensor from operators.hello_operator import HelloOperator default_args = { 'owner': 'airflow', 'depends_on_past': False, 'start_date': datetime(2018, 1, 1), 'email_on_failure': False, 'email_on_retry': False, 'retries': 1, 'retry_delay': timedelta(minutes=5), } with DAG('customdag', max_active_runs=3, schedule_interval='@once', default_args=default_args) as dag: sens = MySensor( task_id='taskA' ) op = MyOperator( task_id='taskB', my_field='some text' ) hello_task = HelloOperator(task_id='sample-task', name='foo_bar') sens >> op >> hello_task
Example plugins/my_airflow_plugin.py
from airflow.plugins_manager import AirflowPlugin from hooks.my_airflow_hook import * from operators.my_airflow_operator import * from utils.my_utils import * class PluginName(AirflowPlugin): name = 'my_airflow_plugin' hooks = [MyHook] operators = [MyOperator] sensors = [MySensor]

The following examples show each of the import statements needed in the custom plugin files.

Example hooks/my_airflow_hook.py
from airflow.hooks.base_hook import BaseHook class MyHook(BaseHook): def my_method(self): print("Hello World")
Example sensors/my_airflow_sensor.py
from airflow.sensors.base_sensor_operator import BaseSensorOperator from airflow.utils.decorators import apply_defaults class MySensor(BaseSensorOperator): @apply_defaults def __init__(self, *args, **kwargs): super(MySensor, self).__init__(*args, **kwargs) def poke(self, context): return True
Example operators/my_airflow_operator.py
from airflow.operators.bash_operator import BaseOperator from airflow.utils.decorators import apply_defaults from hooks.my_hook import MyHook class MyOperator(BaseOperator): @apply_defaults def __init__(self, my_field, *args, **kwargs): super(MyOperator, self).__init__(*args, **kwargs) self.my_field = my_field def execute(self, context): hook = MyHook('my_conn') hook.my_method()
Example operators/hello_operator.py
from airflow.models.baseoperator import BaseOperator from airflow.utils.decorators import apply_defaults class HelloOperator(BaseOperator): @apply_defaults def __init__( self, name: str, **kwargs) -> None: super().__init__(**kwargs) self.name = name def execute(self, context): message = "Hello {}".format(self.name) print(message) return message

Follow the steps in Testing custom plugins using the Amazon MWAA CLI utility, and then Creating a plugins.zip file to zip the contents within your plugins directory. For example, cd plugins.

Creating a plugins.zip file

The following steps describe the steps we recommend to create a plugins.zip file locally.

Step one: Test custom plugins using the Amazon MWAA CLI utility

  • The command line interface (CLI) utility replicates an Amazon Managed Workflows for Apache Airflow environment locally.

  • The CLI builds a Docker container image locally that’s similar to an Amazon MWAA production image. This allows you to run a local Apache Airflow environment to develop and test DAGs, custom plugins, and dependencies before deploying to Amazon MWAA.

  • To run the CLI, see the aws-mwaa-local-runner on GitHub.

Step two: Create the plugins.zip file

You can use a built-in ZIP archive utility, or any other ZIP utility (such as 7zip) to create a .zip file.

Note

The built-in zip utility for Windows OS may add subfolders when you create a .zip file. We recommend verifying the contents of the plugins.zip file before uploading to your Amazon S3 bucket to ensure no additional directories were added.

  1. Change directories to your local Airflow plugins directory. For example:

    myproject$ cd plugins
  2. Run the following command to ensure that the contents have executable permissions (macOS and Linux only).

    plugins$ chmod -R 755 .
  3. Zip the contents within your plugins folder.

    plugins$ zip -r plugins.zip .

Uploading plugins.zip to Amazon S3

You can use the Amazon S3 console or the Amazon Command Line Interface (Amazon CLI) to upload a plugins.zip file to your Amazon S3 bucket.

Using the Amazon CLI

The Amazon Command Line Interface (Amazon CLI) is an open source tool that enables you to interact with Amazon services using commands in your command-line shell. To complete the steps on this page, you need the following:

To upload using the Amazon CLI
  1. In your command prompt, navigate to the directory where your plugins.zip file is stored. For example:

    cd plugins
  2. Use the following command to list all of your Amazon S3 buckets.

    aws s3 ls
  3. Use the following command to list the files and folders in the Amazon S3 bucket for your environment.

    aws s3 ls s3://YOUR_S3_BUCKET_NAME
  4. Use the following command to upload the plugins.zip file to the Amazon S3 bucket for your environment.

    aws s3 cp plugins.zip s3://YOUR_S3_BUCKET_NAME/plugins.zip

Using the Amazon S3 console

The Amazon S3 console is a web-based user interface that allows you to create and manage the resources in your Amazon S3 bucket.

To upload using the Amazon S3 console
  1. Open the Environments page on the Amazon MWAA console.

  2. Choose an environment.

  3. Select the S3 bucket link in the DAG code in S3 pane to open your storage bucket on the Amazon S3 console.

  4. Choose Upload.

  5. Choose Add file.

  6. Select the local copy of your plugins.zip, choose Upload.

Installing custom plugins on your environment

This section describes how to install the custom plugins you uploaded to your Amazon S3 bucket by specifying the path to the plugins.zip file, and specifying the version of the plugins.zip file each time the zip file is updated.

Specifying the path to plugins.zip on the Amazon MWAA console (the first time)

If this is the first time you're uploading a plugins.zip to your Amazon S3 bucket, you also need to specify the path to the file on the Amazon MWAA console. You only need to complete this step once.

  1. Open the Environments page on the Amazon MWAA console.

  2. Choose an environment.

  3. Choose Edit.

  4. On the DAG code in Amazon S3 pane, choose Browse S3 next to the Plugins file - optional field.

  5. Select the plugins.zip file on your Amazon S3 bucket.

  6. Choose Choose.

  7. Choose Next, Update environment.

Specifying the plugins.zip version on the Amazon MWAA console

You need to specify the version of your plugins.zip file on the Amazon MWAA console each time you upload a new version of your plugins.zip in your Amazon S3 bucket.

  1. Open the Environments page on the Amazon MWAA console.

  2. Choose an environment.

  3. Choose Edit.

  4. On the DAG code in Amazon S3 pane, choose a plugins.zip version in the dropdown list.

  5. Choose Next.

Example use cases for plugins.zip

What's next?

  • Test your DAGs, custom plugins, and Python dependencies locally using the aws-mwaa-local-runner on GitHub.