A lot of ECS customers use the Service Connect feature for privately inter-connecting their ECS services, usually their microservices.
When configuring Service Connect for your ECS service, Service Connect will automatically create a Cloud Map service for you if you don't select an existing one. You currently cannot configure a health check for the Cloud Map service when configuring Service Connect.
And when you don't configure a health check during a Cloud Map service creation, AWS Cloud Map will route traffic to service instances (EC2 instances, ECS tasks, EKS pods etc.) whether they are healthy or not.
This causes an issue where traffic from a Service Connect proxy is forwarded to tasks that are in unhealthy states, resulting in failed requests.
You cannot configure your ECS service's Service Connect with a Cloud Map service you created manually.
From the official documentation, AWS Cloud Map recommends configuring custom health checks with a third-party health-checker for your Cloud Map service.
They suggest that the health-checker should update the Cloud Map service with a call to the UpdateInstanceCustomHealthStatus API and update the instance's status to UNHEALTHY in order for Cloud Map to stop routing traffic to that instance.
However, the Cloud Map service created automatically by Service Connect is not configured with any health check, and you cannot update a custom health status of an instance to a Cloud Map service without a health-check configuration.
This application tries to solve the above issues by periodically checking the health and lifecycle state of your ECS tasks and de-registering the tasks from Cloud Map if the health check fails or if the task is in any other state apart from RUNNING.
This is a more bruteforce way of telling Cloud Map to stop routing traffic to an ECS task that is not healthy. Since you cannot configure a Route 53 health check for private Cloud Map namespaces, this application does the health check for you and updates Cloud Map on the same.
I am a Golang newbie, any suggestions and recommendations to improve this program will be highly appreciated!