airflow-remote-jupyter-notebook 0.0.3

Airflow plugin to execute Jupyter Notebook remotely

Airflow run Jupyter Notebook Remote

• Airflow run Jupyter Notebook Remote
  • What is it?
  • Would you mind buying me a coffee?
  • Dependencies
  • Installation
    • Via Pypi Package:
    • Manually
  • Airfow plugin dependencies
  • Test dependences
  • How to contribute
  • Credits
  • Run remote jupyter notebook using Airflow
  • Plugin Usage
  • Run tests

What is it?

This plugin is designed to allow the execution of Jupyter Notebooks remotely from within an Airflow DAG. By using the plugin, users can integrate and manage Jupyter Notebook workflows as part of their Airflow pipelines, ensuring that data analysis or machine learning code can be orchestrated and run automatically within the DAG scheduling system.

The plugin utilizes the Jupyter API to communicate with a Jupyter server, allowing for operations such as starting a kernel, running notebook cells, and managing sessions. It supports both HTTP requests for session and kernel management and WebSocket connections for sending code to execute inside the notebooks.

Package link: https://pypi.org/project/airflow-remote-jupyter-notebook/

Would you mind buying me a coffee?

If you find this library helpful, consider buying me a coffee! Your support helps maintain and improve the project, allowing me to dedicate more time to developing new features, fixing bugs, and providing updates.

Dependencies

• Python 3
• Requests
• Websockets
• Asyncio

Installation

Via Pypi Package:

$ pip install airflow-remote-jupyter-notebook

Manually

# run docker-compose to up Airfow and Jupyter Notebook containers
$ docker-compose up

Airfow plugin dependencies

• Look at requirements.txt

Test dependences

• pytest

How to contribute

Please report bugs and feature requests at https://github.com/marcelo225/airflow-remote-jupyter-notebook/issues

Credits

Lead Developer - Marcelo Vinicius

Run remote jupyter notebook using Airflow

# in root project folder
$ docker-compose up

• Open http://localhost:8080 in your web browser to open Airflow
• Open http://localhost:8888 in your web browser to open Jupyter Notebook, when you need it
• Run test_dag in Airflow

Plugin Usage

from jupyter_plugin.plugin import JupyterDAG # <--------- How to import this plugin
from airflow.models import Variable
import datetime

with JupyterDAG(
    'test_dag',     
    jupyter_url=Variable.get('jupyter_url'),
    jupyter_token=Variable.get('jupyter_token'),
    jupyter_base_path=Variable.get('jupyter_base_path'),
    max_active_runs=1,
    default_args={
        'owner': 'Marcelo Vinicius',
        'depends_on_past': False,
        'start_date': datetime.datetime(2021, 1, 1),
        'email_on_failure': False,
        'email_on_retry': False,
        'retries': 2        
    },
    description=f'DAG test to run some remote Jupyter Notebook file.',
    schedule=2,
    catchup=False
) as dag:

    test1 = dag.create_jupyter_remote_operator(task_id="test1", notebook_path=f"notebooks/test1.ipynb")
    test2 = dag.create_jupyter_remote_operator(task_id="test2", notebook_path=f"notebooks/test2.ipynb")
    test3 = dag.create_jupyter_remote_operator(task_id="test3", notebook_path=f"notebooks/test3.ipynb")

test1 >> test2 >> test3

DAG Attributes	Description
jupyter_url	Jupyter URL server with HTTP or HTTPS
jupyter_token	Jupyter Authentication Token
jupyter_base_path	Base path where your Jupyter notebooks are stored

Task Creation	Explanation
create_jupyter_remote_operator	Method from the JupyterDAG class that creates a task to execute a specified Jupyter notebook on a remote server.
task_id	A unique identifier for the task, used for tracking and logging within Airflow.
notebook_path	Specifies the path to the Jupyter notebook to be executed, relative to the base path.

Run tests

To test the scripts within the Airflow environment, you can use the following command. This will run all tests located in the /home/airflow/tests directory inside the container:

$ docker-compose exec airflow pytest /home/airflow/tests