A payments API runs on Amazon ECS behind an Application Load Balancer. Each release replaces every task at once, and a faulty build has repeatedly caused outages for customers. The team wants to limit the blast radius of bad releases. What should it change?
Show answer and explanation
Answer: A
Staged and blue/green releases either shift traffic gradually or keep the previous version available, so a bad build affects only part of the fleet or can be reverted quickly. Automated health probes keep failing tasks from receiving full traffic, and automatic reversion restores the prior version when alarms fire. Releasing more often with the same all-at-once method simply multiplies the outages.
Why the other options are wrong
- B. More frequent all-at-once releases keep the same blast radius on every deploy, so customers still lose the whole service whenever a build is broken.
- C. Load balancer health checks are what keep unhealthy tasks out of rotation. Removing them lets failing tasks serve traffic and makes the outages worse.
- D. Deploying from personal workstations bypasses review, testing and audit trails, so it increases risk rather than reducing it.