Apache Airflow (or simply Airflow) is a platform to programmatically author, schedule, and monitor workflows. It can be used to author workflows as Directed Acyclic Graphs (DAGs) of tasks and define an automated pipeline of tasks to execute one after the other.
Here is a quick step by step guide on how to use Airflow alongside Spark to automatically run a workflow on Qarnot.
If you are interested in another version, please send us an email at email@example.com.
Before starting a calculation with the Python SDK, a few steps are required:
This tutorial will showcase how to run an Airflow workflow on Qarnot from your computer. The workflow is as follows:
- Start a Spark cluster
- Submit a first Spark app to the cluster: it counts the number of words in the Iliad
- Submit a second Spark app to the cluster: it counts the number of words in the Iliad concatenated 100 times
- Fetch the output of both Spark apps
- Stop the Spark cluster
All these steps will be run in succession without any manual intervention from the user.
Before moving forward, you should setup your working environment to contain the following files which can be downloaded here:
config: contains qarnot and logging config files
custom_operators: scripts developed by Qarnot and needed for our Airflow workflow
dags: contains the script that defines the dag that we will run
logs: directory where the Python and Airflow logs will be stored
spark-resources: text input files of the Iliad and the Iliad concatenated 100 times
Launching the test case
Once you have downloaded all the necessary files, follow the steps below to ensure you have everything you need.
Activate your Python virtual environment and make sure that the Qarnot SDK is installed in it. If you are unsure as to how to do that, you can check the SDK installation documentation for simple steps to follow.
Then install Airflow and its dependencies by running the following command.
pip install apache-airflow['cncf.kubernetes']
Add your ssh public key instead of
<<<MY PUBLIC SSH KEY>>> in
Add your secret token in
config/qarnot.conf instead of
custom_operators/ to your virtual environment site-packages. Make sure to replace
<<<VENV>>> with your virtual environment’s name and the
pythonX with your python version. Note that it is recommended to work inside Python virtual envs to guarantee reproducibility and keep working environments clean.
mv custom_operators/ <<<VENV>>>/lib/pythonX/site-packages/
Set your Airflow home as your current directory
airflow db init
Build your dag named my_first_dag
Launch Airflow workflow on Qarnot with the start date you want
airflow dags backfill my_first_dag -s 2000-01-01
A few notes to keep in mind:
- If you want to launch Airflow a second time, you can add the
--reset-dagrunsflag to bypass some conflicts related to the previous run.
- It is also possible to download input files from a GCP bucket and upload your results to it. It has been excluded from this tutorial for simplicity’s sake. If you are interested in trying it please contact firstname.lastname@example.org.
At any given time, you can monitor the status of your task on the Console as well as from your local terminal.
You can view the outputs in your results bucket
airflow-spark-out. Where you will find the number of words contained in both the Iliad and the version that is concatenated 100 times, as well as different execution logs as shown below.
That’s it! If you have any questions, please contact email@example.com and we will help you with pleasure!