AWS re:Invent 2025 - Accelerate software delivery with Amazon ECS (CNS315)

Accelerating Software Delivery with Amazon ECS Advanced Deployments

Overview

• Presentation covered advanced deployment strategies in Amazon ECS, including benefits, technical details, and real-world use cases.
• Key focus was on how ECS customers can leverage advanced deployment features to improve delivery velocity while maintaining high availability and reliability.

ECS Deployments Primer

• ECS services are defined by two key components:
  • Task definition: Defines a unit of work, a single task within a service
  • Service configuration: Specifies how to deploy one or more tasks to perform a function
• The process of going from initial service revision to the next is an ECS deployment.

Deployment Controllers and Strategies

• ECS provides three deployment controller options:
  1. ECS native controller (focus of this presentation)
  2. CodeDeploy
  3. External (custom deployment management)
• ECS supports four deployment strategies:
  1. Rolling (default)
  2. Blue/Green
  3. Canary
  4. Linear

Advanced Deployment Lifecycle

1. Pre-Scale Up: Allows running custom logic (e.g. admission checks) before scaling up new tasks.
2. Scale Up: Scales up the new "green" environment.
3. Post-Scale Up: No traffic shifted yet, both "blue" and "green" environments running.
4. Test Traffic Shift: Shifts test traffic to the new "green" environment.
5. Post-Test Traffic Shift: Allows running tests/checks on the "green" environment.
6. Production Traffic Shift: Shifts production traffic to the "green" environment.
   • Can be done all at once (Blue/Green), in two stages (Canary), or gradually (Linear).
7. Post-Production Traffic Shift: "Blue" environment kept running for a configurable "bake time" in case rollback is needed.
8. Clean Up: "Blue" environment scaled down, "green" becomes the new "blue" for next deployment.

Deployment Strategies in Action

• UI and Catalog Service: Used Blue/Green for UI, Canary for Catalog to enable gradual rollout and testing.
  • Leveraged ALB's advanced request routing for path-based routing.
• Catalog Service via Service Connect: Used Canary to gradually shift test traffic, implemented performance testing hooks.
• Notification Service: Headless service pulling from a queue, used Blue/Green to enable quick rollback.
• Video Streaming Service: Used Blue/Green with Network Load Balancer to enable full version migration.

Choosing and Migrating Deployment Strategies

• Easily switch between strategies by updating deployment configuration.
• Key considerations:
  • Speed of rollback (advanced deployments provide faster rollback)
  • Consistency of client experience (Blue/Green)
  • Need for gradual rollout and testing (Canary, Linear)
• Can migrate from CodeDeploy to ECS advanced deployments by updating service definition.

Key Takeaways

• ECS advanced deployments provide flexibility to optimize software delivery for various business and technical requirements.
• Lifecycle hooks enable custom logic and testing at key deployment stages.
• Supports multiple service exposure patterns (ALB, Service Connect, headless, NLB) for diverse architectures.
• Enables gradual rollouts, canary testing, and quick rollbacks to maintain high availability.

Additional Resources

• ECS advanced deployments landing page
• Related session on deployment pipelines with ECS