Skip to content

analyticsworld1/airflow-dbt

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

99 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

airflow-dbt

This is a collection of Airflow operators to provide easy integration with dbt.

from airflow import DAG
from airflow_dbt.operators.dbt_operator import (
    DbtSeedOperator,
    DbtSnapshotOperator,
    DbtRunOperator,
    DbtTestOperator,
    DbtCleanOperator,
)
from airflow.utils.dates import days_ago

default_args = {
  'dir': '/srv/app/dbt',
  'start_date': days_ago(0)
}

with DAG(dag_id='dbt', default_args=default_args, schedule_interval='@daily') as dag:

  dbt_seed = DbtSeedOperator(
    task_id='dbt_seed',
  )

  dbt_snapshot = DbtSnapshotOperator(
    task_id='dbt_snapshot',
  )

  dbt_run = DbtRunOperator(
    task_id='dbt_run',
  )

  dbt_test = DbtTestOperator(
    task_id='dbt_test',
    retries=0,  # Failing tests would fail the task, and we don't want Airflow to try again
  )

  dbt_clean = DbtCleanOperator(
    task_id='dbt_clean',
  )

  dbt_seed >> dbt_snapshot >> dbt_run >> dbt_test >> dbt_clean

Installation

Install from PyPI:

pip install airflow-dbt

It will also need access to the dbt CLI, which should either be on your PATH or can be set with the dbt_bin argument in each operator.

Usage

There are five operators currently implemented:

Each of the above operators accept the following arguments:

  • env
    • If set as a kwarg dict, passed the given environment variables as the arguments to the dbt task
  • profiles_dir
    • If set, passed as the --profiles-dir argument to the dbt command
  • target
    • If set, passed as the --target argument to the dbt command
  • dir
    • The directory to run the dbt command in
  • full_refresh
    • If set to True, passes --full-refresh
  • vars
    • If set, passed as the --vars argument to the dbt command. Should be set as a Python dictionary, as will be passed to the dbt command as YAML
  • models
    • If set, passed as the --models argument to the dbt command
  • exclude
    • If set, passed as the --exclude argument to the dbt command
  • select
    • If set, passed as the --select argument to the dbt command
  • selector
    • If set, passed as the --selector argument to the dbt command
  • dbt_bin
    • The dbt CLI. Defaults to dbt, so assumes it's on your PATH
  • verbose
    • The operator will log verbosely to the Airflow logs
  • warn_error
    • If set to True, passes --warn-error argument to dbt command and will treat warnings as errors

Typically you will want to use the DbtRunOperator, followed by the DbtTestOperator, as shown earlier.

You can also use the hook directly. Typically this can be used for when you need to combine the dbt command with another task in the same operators, for example running dbt docs and uploading the docs to somewhere they can be served from.

Building Locally

To install from the repository: First it's recommended to create a virtual environment:

python3 -m venv .venv

source .venv/bin/activate

Install using pip:

pip install .

Testing

To run tests locally, first create a virtual environment (see Building Locally section)

Install dependencies:

pip install . pytest

Run the tests:

pytest tests/

Code style

This project uses flake8.

To check your code, first create a virtual environment (see Building Locally section):

pip install flake8
flake8 airflow_dbt/ tests/ setup.py

Package management

If you use dbt's package manager you should include all dependencies before deploying your dbt project.

For Docker users, packages specified in packages.yml should be included as part your docker image by calling dbt deps in your Dockerfile.

Amazon Managed Workflows for Apache Airflow (MWAA)

If you use MWAA, you just need to update the requirements.txt file and add airflow-dbt and dbt to it.

Then you can have your dbt code inside a folder {DBT_FOLDER} in the dags folder on S3 and configure the dbt task like below:

dbt_run = DbtRunOperator(
  task_id='dbt_run',
  dbt_bin='/usr/local/airflow/.local/bin/dbt',
  profiles_dir='/usr/local/airflow/dags/{DBT_FOLDER}/',
  dir='/usr/local/airflow/dags/{DBT_FOLDER}/'
)

Templating and parsing environments variables

If you would like to run DBT using custom profile definition template with environment-specific variables, like for example profiles.yml using jinja:

<profile_name>:
  outputs:
    <source>:
      database: "{{ env_var('DBT_ENV_SECRET_DATABASE') }}"
      password: "{{ env_var('DBT_ENV_SECRET_PASSWORD') }}"
      schema: "{{ env_var('DBT_ENV_SECRET_SCHEMA') }}"
      threads: "{{ env_var('DBT_THREADS') }}"
      type: <type>
      user: "{{ env_var('USER_NAME') }}_{{ env_var('ENV_NAME') }}"
  target: <source>

You can pass the environment variables via the env kwarg parameter:

import os
...

dbt_run = DbtRunOperator(
  task_id='dbt_run',
  env={
    'DBT_ENV_SECRET_DATABASE': '<DATABASE>',
    'DBT_ENV_SECRET_PASSWORD': '<PASSWORD>',
    'DBT_ENV_SECRET_SCHEMA': '<SCHEMA>',
    'USER_NAME': '<USER_NAME>',
    'DBT_THREADS': os.getenv('<DBT_THREADS_ENV_VARIABLE_NAME>'),
    'ENV_NAME': os.getenv('ENV_NAME')
  }
)

License & Contributing

GoCardless ♥ open source. If you do too, come join us.

About

Apache Airflow integration for dbt

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%