Airflow, a popular OSS workflow orchestration platform, is often used by data scientists and engineers to schedule data pipelines. However, in this post, Jet Marsical describes how Cloudflare uses Airflow to implement “SOP-as-Code” practices. To automate the process whereby new servers are connected to their network, Cloudflare developed a Provisioning-as-a-Service platform using Airflow. Each step from their Standard Operating Procedure (SOP) is implemented as a task in the Airflow DAG, although they also use sensors to block tasks when a human operator may need to intervene. The Airflow tasks provide failure handling, logging, notifications, jinja templating, and macros. Mariscal also explains how they used separable DAGs with well-defined dependencies to construct complex automation workflows.