Manage Airflow connections

Cloud Composer 3 | Cloud Composer 2 | Cloud Composer 1

This page describes how to manage Airflow connections in your environment and access them from your DAGs.

About Airflow connections

Aiflow connections store credentials and other connection information, such as user names, connections strings, and passwords. Your DAGs use connections to communicate and access resources in Google Cloud and other services from your DAGs.

Airflow operators in your DAGs either use a default connection for the operator, or you specify a custom connection name.

About connection security

Most Airflow operators do not accept credentials directly. Instead, they use Airflow connections.

When you create a new environment, Cloud Composer generates a unique, permanent fernet key for the environment and secures connection extras by default. You can view the fernet_key in the Configuration page in Airflow UI.

For more information about how connections and passwords are secured in Airflow, see Securing Connections and Masking sensitive data in the Airflow documentations.

About connection types

Airflow uses connections of different types to connect to specific services. For example, the Google Cloud connection type connects to other services in Google Cloud. As another example, S3 connection type connects to an Amazon S3 bucket.

To add a connection type to Airflow, install a PyPI package with that connection type. Some packages are preinstalled in your environment. For example, you can use connection from the apache-airflow-providers-google package without installing custom PyPI packages.

Preconfigured connections

Cloud Composer configures the following default connections in your environment. You can use these connections to access resources in your project without configuring them.

  • google_cloud_default
  • bigquery_default
  • google_cloud_datastore_default
  • google_cloud_storage_default

Add a connection in Secret Manager

You can store a connection in Secret Manager, without adding it to Airflow. We recommend to use this approach when storing credentials and other sensitive information.

To add a connection in Secret Manager:

  1. Configure Secret Manager for your environment.

  2. Add a secret with the name that matches the pattern for connections.

    For example: airflow-connections-example_connection. In your DAGs, use the connection name without the prefix: example_connection.

  3. Add parameters for the connection:

    JSON format

    Add the JSON representation of your connection as the value of the secret. For example:

    {
  "conn_type": "mysql",
  "host": "example.com",
  "login": "login",
  "password": "password",
  "port": "9000"
}

    For more information about the JSON connection format, see Airflow documentation.

    URI format

    Add the URI representation of your connection as the value of the secret:

    • The secret must store a URI representation of the connection. For example, mysql://login:password@example.com:9000.

    • The URI must be URL-encoded. For example, a password that has a space symbol in it must be URL-encoded as follows: mysql://login:secret%20password@example.com:9000.

    Airflow has a convenience method for generating connection URIs. An example of how to encode a complex URL with JSON extras is available in the Airflow documentation.

  4. Check that all connection parameters are correctly read from Secret Manager.

Add a connection in Airflow

As an alternative to storing your connections in Secret Manager, you can store them in Airflow.

To add a connection in Airflow:

Airflow CLI

Run the connections add Airflow CLI command with Google Cloud CLI. For example:

gcloud composer environments run ENVIRONMENT_NAME \
  --location LOCATION \
  connections add -- \
    --conn-type "mysql" \
    --conn-host "example.com" \
    --conn-port "9000" \
    --conn-login "login" \
    --conn-password "password" \
    example_connection

You can also use the --conn-uri argument:

gcloud composer environments run ENVIRONMENT_NAME \
  --location LOCATION \
  connections add -- \
    --conn-uri "mysql://login:password@example.com:9000" \
    example_connection

Replace the following:

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

Airflow UI

Follow the Airflow documentation on creating connections.

Check that Airflow correctly reads a connection

You can run the connections get Airflow CLI command through Google Cloud CLI to check that a connection is read correctly. For example, if you store a connection in Secret Manager, this provides a way to check if all parameters of a connection are read by Airflow from a secret.

gcloud composer environments run ENVIRONMENT_NAME \
    --location LOCATION \
    connections get \
    -- CONNECTION_NAME

Replace:

  • ENVIRONMENT_NAME with the name of the environment.
  • LOCATION with the region where the environment is located.
  • CONNECTION_NAME with the name of the connection. If your connection is stored in Secret Manager, use the connection name without the connection prefix. For example, specify example_connection instead of airflow-connections-example_connection_json.

Example:

gcloud composer environments run example-environment \
    --location us-central1 \
    connections get \
    -- example_connection -o json

Use Airflow connections in your DAGs

This section shows how to access your connection from a DAG.

Use a Secret Manager connection

Use the name of the connection without the prefix. For example, if your secret is named airflow-connections-aws_s3, specify aws_s3.

transfer_dir_from_s3 = S3ToGCSOperator(
    task_id='transfer_dir_from_s3',
    aws_conn_id='aws_s3',
    prefix='data-for-gcs',
    bucket='example-s3-bucket-transfer-operators',
    dest_gcs='gs://us-central1-example-environ-361f4221-bucket/data/from-s3/')

If you store a default connection in Secret Manager you can omit the connection name. See Airflow documentation for a specific operator to get the default connection name used by an operator. For example, the S3ToGCSOperator Airflow operator uses the aws_default connection by default. You can store this default connection in a secret named airflow-connections-aws_default.

Use a connection stored in Airflow

Use the name of the connection, as it is defined in Airflow:

transfer_dir_from_s3 = S3ToGCSOperator(
    task_id='transfer_dir_from_s3',
    aws_conn_id='aws_s3',
    prefix='data-for-gcs',
    bucket='example-s3-bucket-transfer-operators',
    dest_gcs='gs://us-central1-example-environ-361f4221-bucket/data/from-s3/')

To use the default connection for an operator, omit the connection name. See Airflow documentation for a specific operator to get the default connection name used by an operator. For example, the S3ToGCSOperator Airflow operator uses the aws_default connection by default.

Troubleshooting

If your environment cannot access the secret stored in Secret Manager:

  1. Make sure that Secret Manager is configured in your environment.

  2. Check that connection's name in Secret Manager corresponds to the connection used by Airflow. For example, for a connection named example_connection, the secret name is airflow-connections-example_connection.

  3. Check that Airflow correctly reads a connection.

What's next