What it is: How new container revisions reach production on ECS — the resilient-deployment core for container workloads.

Exam-testable facts:

  • Deployment controllers: ECS (default), CODE_DEPLOY (blue/green powered by CodeDeploy), EXTERNAL (bring your own controller, e.g., a third-party progressive-delivery tool).
  • ECS-controller strategies: ROLLING (default; governed by minimumHealthyPercent/maximumPercent), BLUE_GREEN, LINEAR (equal-percentage traffic steps), CANARY (small slice first, then the rest at once).
  • Deployment circuit breaker: watches a rolling deployment and automatically rolls back on repeated task-launch failures — the “self-healing deploy” answer. Can also roll back on CloudWatch alarms.
  • CODE_DEPLOY blue/green: CodeDeploy stands up a replacement task set, validates it against a test listener, then shifts production traffic (all-at-once, canary, or linear); AppSpec file plus Lambda lifecycle hooks (BeforeAllowTraffic, AfterAllowTraffic).
  • Scaling knobs are different things: Service Auto Scaling changes the task count (target tracking on CPU/memory/ALB request count); for the EC2 launch type, capacity providers with managed scaling change the instance count.
  • Fargate note: with CODE_DEPLOY or EXTERNAL controllers on Fargate, minimumHealthyPercent is ignored.

Common traps:

  • “ECS deployment failed and rolled back by itself” → circuit breaker.
  • Blue/green with test-listener validation → CODE_DEPLOY controller, not the ECS rolling strategy.
  • Don’t confuse task-count scaling (service) with instance-count scaling (capacity provider).

Sources: http://docs.aws.amazon.com/AmazonECS/latest/developerguide/service_definition_parameters.html · https://docs.aws.amazon.com/AmazonECS/latest/APIReference/API_DeploymentConfiguration.html · https://docs.aws.amazon.com/whitepapers/latest/introduction-devops-aws/deployment-strategies-matrix.html

Added during the 2026-10-03 review. All prose is original; facts verified against the AWS documentation links in Sources above.