Airflow

An overview of the Airflow integration with Secoda

Airflow Metadata Extracted

Getting Started with Airflow

There are three ways to connect Airflow to Secoda:

For Google Composer or MWAA, you must use the Plugin method to integrate Airflow with Secoda.

Apache (API) method: Secoda uses the Apache Airflow Rest API to pull the metadata from your Airflow instance
Astronomer method: Secoda uses the Astronomer REST API to pull the metadata from your Astronomer instance.
Plugin method: using Secoda's Airflow plugin, you can add callbacks to certain DAGs and tasks so DAG metadata and lineage gets pushed to Secoda.

1. Apache (API)

There are three steps to get started using Airflow with Secoda:

Enable the Airflow REST API
Connect Airflow to Secoda
Whitelist Secoda IP Address

Enable the Airflow REST API

Secoda uses the Airflow stable REST API, which is only available on Airflow v2.0+. For Secoda to retrieve metadata from Airflow, you need to enable your Airflow's REST API. To do this, edit your airflow.cfg file and change the line below to airflow.api.auth.backend.basic_auth\

[api] auth_backend = airflow.api.auth.backend.deny_all

After changing this is what the line should look like:

[api] auth_backend = airflow.api.auth.backend.basic_auth

Connect Airflow to Secoda

After enabling the Airflow REST API, the next step is to connect Secoda:

In the Secoda App, select ‘Connect Integration’ on the Integrations tab
Search for and select ‘Airflow’
Enter your Airflow user and password. This information is kept encrypted.
Click 'Connect'

Security

VPCs keep servers inaccessible to traffic from the internet. With VPC, you’re able to designate specific web servers access to your servers. In this case, you will be whitelisting the Secoda IPs to read from your Airflow instance.

Allow Secoda to access your Airflow API from the Secoda IP Address.

2. Astronomer

There are three steps to getting started with Astronomer

Retrieve your access token
Retrieve the deployment URL
Connect Astronomer to Secoda

Retrieve your access token

Follow the steps in Create a Workspace API token to create your token. Make sure to save the token on creation in order to use it later in this setup.

Retrieve your deployment URL

Your Deployment URL is the host you use to call the Airflow API.

Run the following command to retrieve the URL for your Deployment Airflow UI:

astro deployment inspect -n <deployment_name> metadata.airflow_api_url

Alternatively, you can retrieve your Deployment URL by opening the Airflow UI for your Deployment on Astro and copying the URL of the page up to /home. For example, if the home page of your Deployment Airflow UI is hosted at clq52c95r000208i8c7wahwxt.astronomer.run/dz3uu847/home, your Deployment URL is clq52c95r000208i8c7wahwxt.astronomer.run/dz3uu847.

Connect Astronomer to Secoda

After retrieving the access token and deployment URL, the next step is to connect Secoda:

In the Secoda App, select ‘Connect Integration’ on the Integrations tab
Search for and select ‘Airflow’
Select the "Astronomer" option
Enter your Astronomer deployment URL and access token. This information is kept encrypted.
Click 'Connect'

3. Plugin

This method will push metadata from your Airflow instance to Secoda. Then on an extraction, Secoda will process all the pushed metadata instead of directly pulling from the API.

Create an API Key

Navigate to your workspace settings on Secoda. Click on the “API” to get a list of your API Keys.

Click on “Generate New API Key” button and then write down the generated key. Please keep this key somewhere safe and do not share with others. This key will have the same access to your Secoda workspace as the user who created it.

Create an Airflow Integration

Navigate to the integrations page and create a new Airflow integration. Change the connection type tab to "Plugin" and then click Submit. Copy the integration ID from the URL of the page and write it down.

Install Secoda Airflow Provider

Install secoda-airflow package using pip:

pip install secoda-airflow

After the installation, navigate to your Airflow connections settings and create a new Secoda connection. Fill in the form with these details

Connection ID: secoda_default
Connection Type: Secoda Hook
Host: {Your Secoda Domain}/api/v1/
- The Secoda Domain is the domain you access Secoda with
Password: enter your API key from first step
Extra JSON: enter your integration ID in the following format

{"integration_id": "id from step 2"}

Save the connection. The Secoda provider is set up!

Add Secoda callbacks to your DAGs

Here is an example DAG with Secoda callbacks.

from airflow import DAG
from airflow.providers.snowflake.operators.snowflake import SnowflakeOperator
from airflow.utils.dates import days_ago

from secoda_airflow.callbacks import task_callbacks, dag_callbacks

SNOWFLAKE_CONN_ID = "snowflake"

query = """
CREATE OR REPLACE TABLE raw.test.customers_copy AS
SELECT * FROM raw.test.customers;
"""


default_args = {
    "owner": "airflow",
    "start_date": days_ago(1),
}

with DAG(
    "create_customers_copy_dag",
    default_args=default_args,
    schedule_interval="@daily",
    catchup=False,
    **dag_callbacks,
) as dag:
    run_query = SnowflakeOperator(
        task_id="run_query",
        snowflake_conn_id=SNOWFLAKE_CONN_ID,
        sql=query,
        **task_callbacks,
    )

You can also import individual callbacks from the secoda_airflow.callbacks

dag_success_callback
task_success_callback
dag_failure_callback
task_failure_callback

Run an extraction in Secoda

After your DAGs have run with the new callbacks, please navigate to your Airflow integration in Secoda and run an extraction to process all of the callbacks.

To automate this step, please set your integration schedule to run extractions right after your Airflow runs have finished.

Alternatively, consider adding a run integration API request to the end of your DAG. That way the extraction will run as part of the DAG process. See below for a sample request.

Endpoint: {Your Secoda Domain}/api/v1/databuilder/jobs/

Method: POST

Payload:

{
    "integration_id": "Your Airflow Integration ID",
    "type": "metadata
}

Last updated 10 days ago