Airflow Summit 2023
Join the Airflow community on September 19—21 during the Airflow Summit 2023 conference to learn more about Airflow and share your expertise. Call for papers is now open

Trigger Cloud Composer DAGs with Cloud Functions and Airflow REST API

Cloud Composer 1 | Cloud Composer 2

This page describes how to use Cloud Functions to trigger Cloud Composer DAGs in response to events.

Apache Airflow is designed to run DAGs on a regular schedule, but you can also trigger DAGs in response to events. One way to do this is to use Cloud Functions to trigger Cloud Composer DAGs when a specified event occurs.

The example in this guide runs a DAG every time a change occurs in a Cloud Storage bucket. Changes to any object in a bucket trigger a function. This function makes a request to Airflow REST API of your Cloud Composer environment. Airflow processes this request and runs a DAG. The DAG outputs information about the change.

Before you begin

Check your environment's networking configuration

This solution does not work in Private IP and VPC Service Controls configurations because it is not possible to configure connectivity from Cloud Functions to the Airflow web server in these configurations.

In Cloud Composer 2, you can use another approach: Trigger DAGs using Cloud Functions and Pub/Sub Messages

Enable APIs for your project

Console

Enable the Cloud Composer and Cloud Functions APIs.

Enable the APIs

gcloud

Enable the Cloud Composer and Cloud Functions APIs: 

gcloud services enable cloudfunctions.googleapis.com composer.googleapis.com

Enable the Airflow REST API

Depending on your version of Airflow:

Allow API calls to Airflow REST API using Webserver Access Control

Cloud Functions can reach out to Airflow REST API either using IPv4 or IPv6 address.

If you are not sure what will be the calling IP range then use a default configuration option in Webserver Access Control which is All IP addresses have access (default) to not accidentally block your Cloud Functions.

Create a Cloud Storage bucket

This example triggers a DAG in response to changes in a Cloud Storage bucket. create a new bucket to use in this example.

Get the Airflow web server URL

This example makes REST API requests to the Airflow web server endpoint. You use the part of the Airflow web interface URL before .appspot.com in your Cloud Function code.

Console

  1. In the Google Cloud console, go to the Environments page.

    Go to Environments

  2. Click the name of your environment.

  3. On the Environment details page, go to the Environment configuration tab.

  4. The URL of the Airflow web server is listed in the Airflow web UI item.

gcloud

Run the following command:

gcloud composer environments describe ENVIRONMENT_NAME \
    --location LOCATION \
    --format='value(config.airflowUri)'

Replace:

  • ENVIRONMENT_NAME with the name of the environment.
  • LOCATION with the region where the environment is located.

Get the client_id of the IAM proxy

To make a request to the Airflow REST API endpoint, the function requires the client ID of the Identity and Access Management proxy that protects the Airflow web server.

Cloud Composer does not provide this information directly. Instead, make an unauthenticated request to the Airflow web server and capture the client ID from the redirect URL:

cURL 

curl -v AIRFLOW_URL 2>&1 >/dev/null | grep -o "client_id\=[A-Za-z0-9-]*\.apps\.googleusercontent\.com"

Replace AIRFLOW_URL with the URL of the Airflow web interface.

In the output, search for the string following client_id. For example:

client_id=836436932391-16q2c5f5dcsfnel77va9bvf4j280t35c.apps.googleusercontent.com

Python

Save the following code in a file called get_client_id.py. Fill in your values for project_id, location, and composer_environment, then run the code in Cloud Shell or your local environment. 

# This script is intended to be used with Composer 1 environments
# In Composer 2, the Airflow Webserver is not in the tenant project
# so there is no tenant client ID
# See https://cloud.google.com/composer/docs/composer-2/environment-architecture
# for more details
import google.auth
import google.auth.transport.requests
import requests
import six.moves.urllib.parse

# Authenticate with Google Cloud.
# See: https://cloud.google.com/docs/authentication/getting-started
credentials, _ = google.auth.default(
    scopes=["https://www.googleapis.com/auth/cloud-platform"]
)
authed_session = google.auth.transport.requests.AuthorizedSession(credentials)

# project_id = 'YOUR_PROJECT_ID'
# location = 'us-central1'
# composer_environment = 'YOUR_COMPOSER_ENVIRONMENT_NAME'

environment_url = (
    "https://composer.googleapis.com/v1beta1/projects/{}/locations/{}"
    "/environments/{}"
).format(project_id, location, composer_environment)
composer_response = authed_session.request("GET", environment_url)
environment_data = composer_response.json()
composer_version = environment_data["config"]["softwareConfig"]["imageVersion"]
if "composer-1" not in composer_version:
    version_error = ("This script is intended to be used with Composer 1 environments. "
                     "In Composer 2, the Airflow Webserver is not in the tenant project, "
                     "so there is no tenant client ID. "
                     "See https://cloud.google.com/composer/docs/composer-2/environment-architecture for more details.")
    raise (RuntimeError(version_error))
airflow_uri = environment_data["config"]["airflowUri"]

# The Composer environment response does not include the IAP client ID.
# Make a second, unauthenticated HTTP request to the web server to get the
# redirect URI.
redirect_response = requests.get(airflow_uri, allow_redirects=False)
redirect_location = redirect_response.headers["location"]

# Extract the client_id query parameter from the redirect.
parsed = six.moves.urllib.parse.urlparse(redirect_location)
query_string = six.moves.urllib.parse.parse_qs(parsed.query)
print(query_string["client_id"][0])

Upload a DAG to your environment

Upload a DAG to your environment. The following example DAG outputs the received DAG run configuration. You trigger this DAG from a function, which you create later in this guide.

import datetime

import airflow
from airflow.operators.bash_operator import BashOperator


with airflow.DAG(
        'composer_sample_trigger_response_dag',
        start_date=datetime.datetime(2021, 1, 1),
        # Not scheduled, trigger only
        schedule_interval=None) as dag:

    # Print the dag_run's configuration, which includes information about the
    # Cloud Storage object change.
    print_gcs_info = BashOperator(
        task_id='print_gcs_info', bash_command='echo {{ dag_run.conf }}')

Deploy a Cloud Function that triggers the DAG

You can deploy a Cloud Function using your preferred language supported by Cloud Functions or Cloud Run. This tutorial demonstrates a Cloud Function implemented in Python and Java.

Specify Cloud Function configuration parameters

  • Trigger. For this example, select a trigger that works when a new object is created in a bucket, or an existing object gets overwritten.

    • Trigger Type. Cloud Storage.

    • Event Type. Finalize / Create.

    • Bucket. Select a bucket that must trigger this function.

    • Retry on failure. We recommend to disable this option for the purposes of this example. If you use your own function in a production environment, enable this option to handle transient errors.

  • Runtime service account, in the Runtime, build, connections and security settings section. Use one of the following options, depending on your preferences:

    • Select Compute Engine default service account. With default IAM permissions, this account can run functions that access Cloud Composer environments.

    • Create a custom service account that has the Composer User role and specify it as a runtime service account for this function. This option follows the minimum privilege principle.

  • Runtime and entry point, on the Code step. When adding code for this example, select the Python 3.7 or later runtime and specify trigger_dag as the entry point.

Add requirements

Specify the dependencies in the requirements.txt file:

requests-toolbelt==1.0.0
google-auth==2.17.3
google-cloud-pubsub==2.17.0

Put the following code to the main.py file and make the following replacements:

  • Replace the value of the client_id variable with the client_id value that you obtained earlier.

  • Replace the value of the webserver_id variable with your tenant project ID, which is a part of the Airflow web interface URL before .appspot.com. You obtained the Airflow web interface URL earlier.

  • Specify the Airflow REST API version that you use:

    • If you use the stable Airflow REST API, set the USE_EXPERIMENTAL_API variable to False.
    • If you use the experimental Airflow REST API, no changes are needed. The USE_EXPERIMENTAL_API variable is already set to True.


from google.auth.transport.requests import Request
from google.oauth2 import id_token
import requests


IAM_SCOPE = 'https://www.googleapis.com/auth/iam'
OAUTH_TOKEN_URI = 'https://www.googleapis.com/oauth2/v4/token'
# If you are using the stable API, set this value to False
# For more info about Airflow APIs see https://cloud.google.com/composer/docs/access-airflow-api
USE_EXPERIMENTAL_API = True


def trigger_dag(data, context=None):
    """Makes a POST request to the Composer DAG Trigger API

    When called via Google Cloud Functions (GCF),
    data and context are Background function parameters.

    For more info, refer to
    https://cloud.google.com/functions/docs/writing/background#functions_background_parameters-python

    To call this function from a Python script, omit the ``context`` argument
    and pass in a non-null value for the ``data`` argument.

    This function is currently only compatible with Composer v1 environments.
    """

    # Fill in with your Composer info here
    # Navigate to your webserver's login page and get this from the URL
    # Or use the script found at
    # https://github.com/GoogleCloudPlatform/python-docs-samples/blob/main/composer/rest/get_client_id.py
    client_id = 'YOUR-CLIENT-ID'
    # This should be part of your webserver's URL:
    # {tenant-project-id}.appspot.com
    webserver_id = 'YOUR-TENANT-PROJECT'
    # The name of the DAG you wish to trigger
    dag_name = 'composer_sample_trigger_response_dag'

    if USE_EXPERIMENTAL_API:
        endpoint = f'api/experimental/dags/{dag_name}/dag_runs'
        json_data = {'conf': data, 'replace_microseconds': 'false'}
    else:
        endpoint = f'api/v1/dags/{dag_name}/dagRuns'
        json_data = {'conf': data}
    webserver_url = (
        'https://'
        + webserver_id
        + '.appspot.com/'
        + endpoint
    )
    # Make a POST request to IAP which then Triggers the DAG
    make_iap_request(
        webserver_url, client_id, method='POST', json=json_data)


# This code is copied from
# https://github.com/GoogleCloudPlatform/python-docs-samples/blob/main/iap/make_iap_request.py
# START COPIED IAP CODE
def make_iap_request(url, client_id, method='GET', **kwargs):
    """Makes a request to an application protected by Identity-Aware Proxy.
    Args:
      url: The Identity-Aware Proxy-protected URL to fetch.
      client_id: The client ID used by Identity-Aware Proxy.
      method: The request method to use
              ('GET', 'OPTIONS', 'HEAD', 'POST', 'PUT', 'PATCH', 'DELETE')
      **kwargs: Any of the parameters defined for the request function:
                https://github.com/requests/requests/blob/master/requests/api.py
                If no timeout is provided, it is set to 90 by default.
    Returns:
      The page body, or raises an exception if the page couldn't be retrieved.
    """
    # Set the default timeout, if missing
    if 'timeout' not in kwargs:
        kwargs['timeout'] = 90

    # Obtain an OpenID Connect (OIDC) token from metadata server or using service
    # account.
    google_open_id_connect_token = id_token.fetch_id_token(Request(), client_id)

    # Fetch the Identity-Aware Proxy-protected URL, including an
    # Authorization header containing "Bearer " followed by a
    # Google-issued OpenID Connect token for the service account.
    resp = requests.request(
        method, url,
        headers={'Authorization': 'Bearer {}'.format(
            google_open_id_connect_token)}, **kwargs)
    if resp.status_code == 403:
        raise Exception('Service account does not have permission to '
                        'access the IAP-protected application.')
    elif resp.status_code != 200:
        raise Exception(
            'Bad response from application: {!r} / {!r} / {!r}'.format(
                resp.status_code, resp.headers, resp.text))
    else:
        return resp.text
# END COPIED IAP CODE

Test your function

To check that your function and DAG work as intended:

  1. Wait until your function deploys.
  2. Upload a file to your Cloud Storage bucket. As an alternative, you can trigger the function manually by selecting the Test the function action for it in Google Cloud console.
  3. Check the DAG page in the Airflow web interface. The DAG should have one active or already completed DAG run.
  4. In the Airflow UI, check task logs for this run. You should see that the print_gcs_info task outputs the data received from the function to the logs:
[2021-04-04 18:25:44,778] {bash_operator.py:154} INFO - Output:
[2021-04-04 18:25:44,781] {bash_operator.py:158} INFO - Triggered from GCF:
    {bucket: example-storage-for-gcf-triggers, contentType: text/plain,
    crc32c: dldNmg==, etag: COW+26Sb5e8CEAE=, generation: 1617560727904101,
    ... }
[2021-04-04 18:25:44,781] {bash_operator.py:162} INFO - Command exited with
    return code 0h

What's next