I'm setting up a simple Hello World ECS cluster using EC2 instances through a capacity provider. The capacity provider and Auto Scaling Group appear healthy, the EC2 instance is registered with the cluster, the ECS agent container is running, and the task's CPU and memory requirements fit the instance. However, every task remains stuck in a provisioning loop and eventually stops with `TaskFailedToStart: ATTRIBUTE`.
The instance is using an ECS-optimized AMI that appears outdated, with Amazon ECS Agent version 1.51.0. The console also warns that one or more container instances are not running the latest ECS agent. Could the old AMI or agent version be causing the missing attribute, and where should I find the current ECS-optimized AMI?
1 Answer
You generally shouldn't need to use the Marketplace search for ECS-optimized images. AWS publishes current ECS-optimized AMIs for each region, and the recommended approach is to retrieve the latest image ID through the public SSM parameter rather than selecting an old catalog listing manually. After launching a new instance with the current AMI, confirm that it registers with the intended cluster and capacity provider.
I found the old image through the EC2 AMI catalog search, where it was the only result labeled as ECS-optimized. I hadn't realized the current recommended images were available through the regional SSM parameters instead.