AWS re:Invent 2025 - Deep Dive: ECS Managed Instances & Blue/Green for Resilient Services (CNS416)

AWS re:Invent 2025 - Deep Dive: ECS Managed Instances & Blue/Green for Resilient Services

Overview of Amazon ECS

• Amazon ECS is a tier-one service that is a prerequisite for any new AWS region
• ECS launches over 3 billion tasks per week across all regions
• Over 65% of new AWS customers start with ECS due to its simplicity and integration with other AWS services
• ECS powers many internal AWS services like Amazon SageMaker, Amazon Lex, and Amazon Polly

ECS Managed Instances

Shared Responsibility Model

• ECS on EC2 allows customers to bring their own EC2 instances and manage the underlying infrastructure
• ECS Fargate is a fully serverless offering where AWS manages all the infrastructure
• ECS Managed Instances aims to provide the flexibility of ECS on EC2 with the simplicity of Fargate

Key Features

• ECS Managed Instances is built on top of EC2 Managed Instances, allowing AWS to manage the EC2 instances on the customer's behalf
• Customers provide two IAM roles - one for the instance and one for infrastructure management
• Instances are automatically provisioned, patched, and replaced by ECS to maintain compliance
• Customers can specify compute requirements at the task definition level, and ECS will provision the appropriate EC2 instances
• ECS uses a "first fit descending" approach to pack tasks onto the largest available instances to optimize cost and throughput

Patching and Host Replacement

• ECS Managed Instances replaces instances every 30 days to apply OS and security patches
• Instances are replaced during customer-defined maintenance windows to minimize disruption
• Unmanaged tasks are the last to be deprovisioned during host replacement to ensure availability

When to Use ECS Managed Instances

• Recommended for customers who need more flexibility than Fargate, such as larger task sizes or specific instance types
• Useful for workloads that can benefit from the elasticity and burst capabilities of larger EC2 instances
• Provides a good balance between the control of ECS on EC2 and the simplicity of Fargate

ECS Deployments

Blue/Green Deployments

• Blue/Green deployments allow customers to deploy a new version of a service in parallel with the current production version
• The deployment process includes preparation, deployment, testing, traffic shift, monitoring, and cleanup phases
• Deployment lifecycle hooks allow customers to run custom validation tests at each phase
• Traffic is instantly shifted from the blue (current) version to the green (new) version when ready

Linear Deployments

• Linear deployments gradually shift traffic from the current (blue) version to the new (green) version over time
• Traffic is shifted in increments (e.g., 25%, 50%, 75%, 100%) with validation checks at each step
• Allows for a more conservative, step-by-step approach to rolling out a new service version

Canary Deployments

• Canary deployments start by directing a small portion (e.g., 10%) of traffic to the new (green) version
• After validation, the remaining traffic is instantly shifted to the new version
• Provides a middle ground between the speed of blue/green and the caution of linear deployments

Choosing the Right Deployment Strategy

• Blue/green is best for fast, high-confidence deployments
• Linear is recommended for more conservative, step-by-step rollouts
• Canary is a compromise, allowing for gradual validation before full rollout

Key Takeaways

• ECS Managed Instances provides the flexibility of EC2 with the simplicity of Fargate, allowing customers to focus on their applications
• ECS deployments offer multiple strategies (blue/green, linear, canary) to suit different application needs and risk profiles
• Deployment lifecycle hooks enable customers to integrate custom validation and testing into the deployment process
• AWS aims to bake in best practices and operational experience into the ECS service to simplify infrastructure management for customers

Additional Resources

• ECS Managed Instances Documentation
• Blog Post: Deep Dive on ECS Managed Instances
• Builder Library: Ensuring Rollback Safety with Deployments
• Builder Library: How We Build Scalable, Resilient Services