DEA-C01 Practice Question 27: MWAA Airflow Migration
AWS Certified Data Engineer – Associate (DEA-C01) · Test 3 · question 7 of 10
A data platform team runs roughly 200 Apache Airflow DAGs on a pair of self-managed Amazon EC2 instances. They spend most of their on-call time patching the hosts, recovering a scheduler that has fallen over, and manually adding workers before month-end runs. They want to keep their DAGs, their Airflow operators, and the Airflow UI that analysts already use, while handing the infrastructure to AWS.
Which option best meets these requirements?
Show answer and explanation
Answer:
Amazon MWAA is a managed service for Apache Airflow: "you can use Apache Airflow and Python to create workflows without managing infrastructure for scalability, availability, and security." MWAA "sets up Apache Airflow for you using the same Apache Airflow user interface and open-source code available on the internet", so the existing DAGs and operators carry over and analysts see the UI they already know. It autoscales workers between limits you set, which covers the month-end spike, and AWS patches and upgrades the environment. DAGs are deployed by putting them in the environment's S3 bucket.
Option A throws away the investment. Two hundred DAGs rewritten in Amazon States Language is a large migration project, the Airflow UI and operator ecosystem are lost, and none of that work is required to get off the EC2 hosts.
Option C is a rewrite as well, into a less expressive model. Glue workflows orchestrate Glue jobs and crawlers with triggers; they are not a general-purpose scheduler and cannot run arbitrary Airflow operators against the wide range of systems these DAGs touch.
Option D keeps Airflow but keeps the operational burden too; arguably increases it. The team would now patch and upgrade Airflow *and* run a Kubernetes cluster, which is the opposite of handing infrastructure to AWS.
Reference: What Is Amazon Managed Workflows for Apache Airflow?