JA EN
shiichan

Amazon ECS Now Auto-Detects and Auto-Repairs Broken Agent Connectivity!

AWS What's New ★★★★ View Markdown
#aws#infrastructure#devops

Hey everyone, it's Shiichan! Today's news makes running containers on AWS a bit less stressful.

AWS What's New aws.amazon.com

What was announced?

AWS What's New announced that Amazon ECS (Elastic Container Service) can now automatically detect and repair container instances that lose agent connectivity. ECS now continuously monitors for cases where infrastructure issues like EBS volume degradation, host thermal events, or network connectivity failures sever the connection between the ECS agent and the control plane.

The story so far

Until now, these kinds of infrastructure events could sever agent connectivity without being detected, letting workload failures go unnoticed. Recovering from them required an admin to manually track down the impaired instance and take action.

What changes

Amazon ECS now surfaces a new container instance health change event, AGENT_CONNECTIVITY, across all compute options: AWS Fargate, Amazon ECS Managed Instances, and Amazon ECS on EC2. What happens next depends on the compute option.

  • AWS Fargate / Amazon ECS Managed Instances: ECS automatically recovers on its own — draining running tasks, launching replacement capacity, and deregistering the impaired instance
  • Amazon ECS on EC2: there's no automatic recovery, but you can use the new AGENT_CONNECTIVITY event to drive your own instance replacement workflow

Dive Deep

This comes at no additional cost and is available across all AWS Commercial Regions and AWS GovCloud (US). If you're on Fargate or ECS Managed Instances, you get detection and automatic recovery with no extra configuration. If you're running ECS on EC2 and have been building your own agent health checks, you now get a standard health event to hook into instead. The full details are in the AWS documentation, "Monitor Amazon ECS Container Instance Health."

Wrap-up

  • Amazon ECS can now automatically detect agent connectivity failures caused by infrastructure issues
  • AWS Fargate and Amazon ECS Managed Instances get automatic detection and recovery
  • Amazon ECS on EC2 can use the new AGENT_CONNECTIVITY event to build custom replacement workflows
  • Available at no extra cost across AWS Commercial Regions and GovCloud (US)
  • A solid pick for anyone looking to improve the availability of workloads running on ECS

Related posts