Skip to main content
. 2019 Jul 18;8(7):giz084. doi: 10.1093/gigascience/giz084

Figure 2:

Figure 2:

Structure diagram for scaling out CWL-Airflow with a Celery cluster of 4 nodes. Node 1 runs the Airflow database to save task metadata and the Airflow scheduler with the Celery executor to submit tasks for processing to the Airflow celery workers on Nodes 2, 3, and 4. The Airflow and Flower (Celery) web servers allow for monitoring and controlling of the task execution process. All nodes have shared access to the dags, jobs, temp, and output folders.