EC2 CloudWatch Metrics Fetch Agent

I need a small, reliable agent that is hosted on agent core runtime and can be involved after every test run for Root cause analysis . The agent connects to the AWS account and routinely pulls CloudWatch data from a ECS cluster that host 15 microservices . The focus is on five metrics—CPU utilization, Memory usage, Request count, Response time, and Error rate—so the solution must fetch both the default statistics and the custom ones that require the CloudWatch Agent or an equivalent approach.

AWS is our only cloud provider, and ECS cluster is the only service in scope, so the code can stay lightweight. I am comfortable running the agent as either a Python script (boto3), a serverless Lambda,


Deliverables
• Source code with clear comments
. walkthrough of the code
• A one-step deployment script or Terraform module that creates any IAM roles/policies the agent needs

• A README that shows how to schedule the collector, adjust polling intervals, and test the output locally

Acceptance criteria
• The agent returns the five named metrics for any chosen ECS instance ID on demand
• walkthrough of the code