Test, synchronize, and deploy your DAGs from GitHub

Cloud Composer 3 | Cloud Composer 2 | Cloud Composer 1

This guide explains how to create a CI/CD pipeline to test, synchronize, and deploy DAGs to your Cloud Composer environment from your GitHub repository.

If you want to only synchronize data from other services, see Transfer data from other services.

CI/CD pipeline overview

Figure 1. Architecture diagram showing the steps of the flow (click to enlarge)

The CI/CD pipeline that to test, synchronize, and deploys DAGs has the following steps:

1. You make a change to a DAG and push that change to a development branch in your repository.

2. You open a pull request against the main branch of your repository.

3. Cloud Build runs unit tests to check that your DAG is valid.

4. Your pull request is approved and merged to the main branch of your repository.

5. Cloud Build synchronizes your development Cloud Composer environment with these new changes.

6. You verify that the DAG behaves as expected in your development environment.

7. If your DAG works as expected, you upload the DAG to your production Cloud Composer environment.

Objectives

Objective:

• Run an automated presubmit check using Cloud Build. This check runs unit tests for a DAG.
• Synchronize DAGs in your development Cloud Composer environment with DAGs in your GitHub repository.

Before you begin

• This guide assumes that you are working with two identical Cloud Composer environments: a development environment and a production environment.

  For the purposes of this guide, you are configuring a CI/CD pipeline only for your development environment. Make sure that the environment you use is not a production environment.

• This guide assumes that you have your DAGs and their tests stored in a GitHub repository.

  The example CI/CD pipeline demonstrates the contents of an example repository. DAGs and tests are stored in the dags/ directory, with requirements files, the constraints file, and Cloud Build configuration files stored at the top level. The DAG synchronization utility and its requirements are located in the utils directory.

Create a presubmit check job and unit tests

The first Cloud Build job runs a presubmit check, which executes unit tests for your DAGs.

Add unit tests

If you have not already, author unit tests for your DAGs. Save these tests alongside the DAGs in your repository, each with the _test suffix. For example, the test file for the DAG in example_dag.py is example_dag_test.py. These are the tests that run as a presubmit check in your repository.

Create Cloud Build YAML configuration for the presubmit check

In your repository, create a YAML file named test-dags.cloudbuild.yaml that configures your Cloud Build job for presubmit checks. In it, there are three steps:

1. Install the dependencies needed by your DAGs.
2. Install the dependencies needed by your unit tests.
3. Execute the DAG tests.

steps:
  # install dependencies
  - name: python:3.8-slim
    entrypoint: pip
    args: ["install", "-r", "requirements.txt", "-c", "constraints.txt", "--user"]

  - name: python:3.8-slim
    entrypoint: pip
    args: ["install", "-r", "requirements-test.txt", "--user"]

  # run in python 3.8 which is latest version in Cloud Composer
  - name: python:3.8-slim
    entrypoint: python3.8
    args: ["-m", "pytest", "-s", "dags/"]

Note: Cloud Build pulls the Python image from Docker Hub and you can specify the tag in the name field of each step. If you are using python as your entrypoint, make sure that your entrypoint matches the version specified in your tag, otherwise entrypoint defaults to latest.

Create the Cloud Build trigger for the presubmit check

Follow the Building repositories from GitHub guide to create a GitHub app based trigger with the following configurations:

• Name: test-dags

• Event: Pull Request

• Source - Repository: choose your repository

• Source - Base branch: ^main$ (change main to the name of your repository's base branch, if required)

• Source - Comment Control: not required

  Caution: Setting Comment Control to be not required means that any user can submit a Pull Request that triggers the build. For larger repositories or repositories with untrusted contributors, requiring Comment Control is recommended.

• Build Configuration - Cloud build configuration file: /test-dags.cloudbuild.yaml (the path to your build file)

Create a DAG sync job and add DAGs utility script

Next, configure a Cloud Build job that runs a DAGs utility script. The utility script in this job synchronizes your DAGs with your Cloud Composer environment after they are merged to the main branch in your repository.

Note: Cloud Composer parses all DAGs on a repeated interval. This happens even if DAGs did not change since the last parse. Because of this behavior, this utility uploads all DAGs to Cloud Composer, not just those that have changed.

Add the DAGs utility script

Add the DAG utility script to your repository. This utility script copies all DAG files in the dags/ directory of your repository to a temporary directory, ignoring all non-DAG Python files. The script then uses the Cloud Storage client library to upload all files from that temporary directory to the dags/ directory in your Cloud Composer environment's bucket.

from __future__ import annotations

import argparse
import glob
import os
from shutil import copytree, ignore_patterns
import tempfile

# Imports the Google Cloud client library
from google.cloud import storage


def _create_dags_list(dags_directory: str) -> tuple[str, list[str]]:
    temp_dir = tempfile.mkdtemp()

    # ignore non-DAG Python files
    files_to_ignore = ignore_patterns("__init__.py", "*_test.py")

    # Copy everything but the ignored files to a temp directory
    copytree(dags_directory, f"{temp_dir}/", ignore=files_to_ignore, dirs_exist_ok=True)

    # The only Python files left in our temp directory are DAG files
    # so we can exclude all non Python files
    dags = glob.glob(f"{temp_dir}/*.py")
    return (temp_dir, dags)


def upload_dags_to_composer(
    dags_directory: str, bucket_name: str, name_replacement: str = "dags/"
) -> None:
    """
    Given a directory, this function moves all DAG files from that directory
    to a temporary directory, then uploads all contents of the temporary directory
    to a given cloud storage bucket
    Args:
        dags_directory (str): a fully qualified path to a directory that contains a "dags/" subdirectory
        bucket_name (str): the GCS bucket of the Cloud Composer environment to upload DAGs to
        name_replacement (str, optional): the name of the "dags/" subdirectory that will be used when constructing the temporary directory path name Defaults to "dags/".
    """
    temp_dir, dags = _create_dags_list(dags_directory)

    if len(dags) > 0:
        # Note - the GCS client library does not currently support batch requests on uploads
        # if you have a large number of files, consider using
        # the Python subprocess module to run gsutil -m cp -r on your dags
        # See https://cloud.google.com/storage/docs/gsutil/commands/cp for more info
        storage_client = storage.Client()
        bucket = storage_client.bucket(bucket_name)

        for dag in dags:
            # Remove path to temp dir
            dag = dag.replace(f"{temp_dir}/", name_replacement)

            try:
                # Upload to your bucket
                blob = bucket.blob(dag)
                blob.upload_from_filename(dag)
                print(f"File {dag} uploaded to {bucket_name}/{dag}.")
            except FileNotFoundError:
                current_directory = os.listdir()
                print(
                    f"{name_replacement} directory not found in {current_directory}, you may need to override the default value of name_replacement to point to a relative directory"
                )
                raise

    else:
        print("No DAGs to upload.")


if __name__ == "__main__":
    parser = argparse.ArgumentParser(
        description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter
    )
    parser.add_argument(
        "--dags_directory",
        help="Relative path to the source directory containing your DAGs",
    )
    parser.add_argument(
        "--dags_bucket",
        help="Name of the DAGs bucket of your Composer environment without the gs:// prefix",
    )

    args = parser.parse_args()

    upload_dags_to_composer(args.dags_directory, args.dags_bucket)

Create Cloud Build YAML configuration for synchronizing DAGs

In your repository, create a YAML file named add-dags-to-composer.cloudbuild.yaml that configures your Cloud Build job for synchronizing DAGs. In it, there are two steps:

1. Install the dependencies needed by the DAGs utility script.

2. Run the utility script to synchronize the DAGs in your repository with your Cloud Composer environment.

steps:
  # install dependencies
  - name: python
    entrypoint: pip
    args: ["install", "-r", "utils/requirements.txt", "--user"]

  # run
  - name: python
    entrypoint: python
    args: ["utils/add_dags_to_composer.py", "--dags_directory=${_DAGS_DIRECTORY}", "--dags_bucket=${_DAGS_BUCKET}"]

Note: Cloud Build pulls the Python image from Docker Hub and you can specify the tag in the name field of each step. If you are using python as your entrypoint, make sure that your entrypoint matches the version specified in your tag, otherwise entrypoint defaults to latest.

Create the Cloud Build trigger

Follow the Building repositories from GitHub guide to create a GitHub app based trigger with the following configurations:

• Name: add-dags-to-composer

• Event: Push to a branch

• Source - Repository: choose your repository

• Source - Base branch: ^main$ (change main to the name of your repository's base branch, if required)

• Source - Included files filter (glob): dags/**

• Build Configuration - Cloud build configuration file: /add-dags-to-composer.cloudbuild.yaml (the path to your build file)

In the Advanced configuration, add two substitution variables:

• _DAGS_DIRECTORY - the directory where dags are located in your repository. If you are using the example repository from this guide, it is dags/.

• _DAGS_BUCKET - the Cloud Storage bucket that contains the dags/ directory in your development Cloud Composer environment. Omit the gs:// prefix. For example: us-central1-example-env-1234ab56-bucket.

Caution: If you do not remove the dags/ suffix from the Cloud Storage bucket name in the _DAGS_BUCKET substitution variable, the DAG sync job fails with a NotFound error. This happens because in Cloud Storage, dags/ is a folder, which is considered part of the object name, not part of the bucket name.

Test your CI/CD pipeline

In this section, follow a DAG development flow that utilizes your newly created Cloud Build triggers.

Run a presubmit job

Create a pull request to your main branch to test your build. Locate your presubmit check on the page. Click Details and choose View more details on Google Cloud Build to see your build logs in the Google Cloud console.

Important: Cloud Build runs unit tests with packages specified in requirements.txt and requirements-test.txt, which include a public version of Airflow. Your Cloud Composer environment might have a different set of PyPI packages and preinstalled packages. We recommend to run a local Airflow environment while developing your DAGs.

Figure 2. Screenshot of the Cloud Build presubmit check status on GitHub (click to enlarge)

If your presubmit check failed, see Addressing build failures.

Validate that your DAG works in your development Cloud Composer environment

After your pull request is approved, merge it to your main branch. Use the Google Cloud console to view your build results. If you have many Cloud Build triggers, you can filter your builds on the trigger name add-dags-to-composer.

After the Cloud Build sync job succeeds, the synchronized DAG appears in your development Cloud Composer environment. There, you can validate that the DAG functions as expected.

Add the DAG to your production environment

After the DAG performs as expected, manually add it to your production environment. To do so, upload the DAG file to the dags/ directory in your production Cloud Composer environment's bucket.

If your DAG sync job failed or if your DAG is not behaving as expected in your development Cloud Composer environment, see Addressing build failures.

Addressing build failures

This section explains how to address common build failure scenarios.

What if my presubmit check failed?

From your pull request, click Details and choose View more details on Google Cloud Build to see your build logs in the Google Cloud console. Use these logs to help you debug the problem with your DAG. Once you have resolved the issues, commit the fix and push to your branch. The presubmit check runs again, and you can continue to iterate using the logs as a debugging tool.

What if my DAG sync job failed?

Use the Google Cloud console to view your build results. If you have many Cloud Build triggers, you can filter your builds on the trigger name add-dags-to-composer. Examine the logs of the build job and resolve the errors. If you need additional help resolving the errors, utilize support channels.

What if my DAG does not work properly in my Cloud Composer environment?

If your DAG does not function as expected in your development Cloud Composer environment, do not manually promote the DAG to your production Cloud Composer environment. Instead, do one of the following:

• Revert the pull request with the changes that broke your DAG to restore it to the state immediately prior to your changes (this also reverts all other files in that pull request).
• Create a new pull request to manually revert changes to the broken DAG.
• Create a new pull request the fix the errors in your DAG.

Following any of these steps triggers a new presubmit check and upon merge, the DAG sync job.

What's next

• Run local Airflow environments
• Write DAGs
• Schedule and trigger DAGs
• Test DAGs