afctl – CLI to manage and deploy Airflow project faster and smoother

afctl

The proposed CLI tool is authored to make creating and deployment of airflow projects faster and smoother. As of now, there is no tool out there that can empower the user to create a boilerplate code structure for airflow projects and make development + deployment of projects seamless.

Requirements

• Python 3.5+
• Docker

Getting Started

1. Installation

Create a new python virtualenv. You can use the following command.

python3 -m venv <name>

Activate your virtualenv

source /path_to_venv/bin/activate

pip3 install afctl

2. Initialize a new afctl project.

The project is created in your present working directory. Along with this a configuration file with the same name is generated in /home/.afctl_configs directory.

afctl init <name of the project>

Eg.

afctl init project_demo

• The following directory structure will be generated

.
├── deployments
│   └── project_demo-docker-compose.yml
├── migrations
├── plugins
├── project_demo
│   ├── commons
│   └── dags
├── requirements.txt
└── tests

If you already have a git repository and want to turn it into an afctl project. Run the following command :-

afctl init .

3. Add a new module in the project.

afctl generate module -n <name of the module>

The following directory structure will be generated :

afctl generate module -n first_module
afctl generate module -n second_module

.
├── deployments
│   └── project_demo-docker-compose.yml
├── migrations
├── plugins
├── project_demo
│   ├── commons
│   └── dags
│       ├── first_module
│       └── second_module
├── requirements.txt
└── tests
    ├── first_module
    └── second_module

4. Generate dag

afctl generate dag -n <name of dag> -m <name of module>

The following directory structure will be generate :

afctl generate dag -n new -m first_module

.
├── deployments
│   └── project_demo-docker-compose.yml
├── migrations
├── plugins
├── project_demo
│   ├── commons
│   └── dags
│       ├── first_module
│       │   └── new_dag.py
│       └── second_module
├── requirements.txt
└── tests
    ├── first_module
    └── second_module

The dag file will look like this :

from airflow import DAG
from datetime import datetime, timedelta

default_args = {
'owner': 'project_demo',
# 'depends_on_past': ,
# 'start_date': ,
# 'email': ,
# 'email_on_failure': ,
# 'email_on_retry': ,
# 'retries': 0

}

dag = DAG(dag_id='new', default_args=default_args, schedule_interval='@once')

5. Deploy project locally

You can add python packages that will be required by your dags in requirements.txt . They will automatically get installed.

• To deploy your project, run the following command (make sure docker is running) :

afctl deploy local

If you do not want to see the logs, you can run

afctl deploy local -d

This will run it in detached mode and won't print the logs on the console.

• You can access your airflow webserver on browser at localhost:8080

6. Deploy project on production

• Here we will be deploying our project to Qubole . Sign up at us.qubole.com.
• add git-origin and access-token (if want to keep the project as private repo on Github) to the configs.
• Push the project once completed to Github.
• Deploying to Qubole will require adding deployment configurations.

afctl config add -d qubole -n <name of deployment> -e <env> -c <cluster-label> -t <auth-token>

This command will modify your config file. You can see your config file with the following command :

afctl config show

For example -

afctl config add -d qubole -n demo -e https://api.qubole.com -c airflow_1102 -t khd34djs3

• To deploy run the following command

afctl deploy qubole -n <name>

The following video also contains all the steps of deploying project using afctl -

https://www.youtube.com/watch?v=A4rcZDGtJME&feature=youtu.be

Manage configurations

The configuration file is used for deployment contains the following information.

global:
-airflow_version:
-git:
--origin:
--access-token:
deployment:
-qubole:
--local:
---compose:

• airflow_version can be added to the project when you initialize the project.

afctl init <name> -v <version>

• global configs (airflow_version, origin, access-token) can all be added/ updated with the following command :

afctl config global -o <git-origin> -t <access-token> -v <airflow_version>

Usage

Commands right now supported are

• init
• config
• deploy
• list
• generate

To learn more, run

afctl <command> -h

Caution

Not yet ported for Windows.

Credits

Docker-compose file : https://github.com/puckel/docker-airflow