How do I troubleshoot issues with Container Insights for Amazon ECS?

Resolution

Container Insights is missing metrics for the Fargate launch type

This issue typically occurs when there aren't enough task resources. To resolve this issue, optimize your task resources.

The awslogs container instance log driver loses logs

The logging server might occasionally be unreachable or can't accept logs. In these scenarios, you might experience log loss. To avoid log loss, configure the logs to use non-blocking mode in the Amazon ECS task definition:

"logConfiguration": {
    "logDriver": "awslogs",
    "options": {
        "mode": "non-blocking",
        "max-buffer-size": "25m",
    }
} 

Important: When you use non-blocking mode, the log driver stores logs in an in-memory intermediate buffer. If the buffer runs out of space, then the buffer can't store additional logs and you lose the logs. Also, log loss can occur when the log driver can't manage the application's log emission rate.

You deployed Container Insights on a deleted and recreated cluster on Amazon ECS

Issues can occur when you delete an existing Amazon ECS cluster that doesn't have Container Insights activated. If you create a new cluster with the same name, then you must manually activate Container Insights for the new cluster.

You deactivated Container Insights in the Amazon ECS cluster, but it's still activated

If Container Insights automatically reactivates, then you must manually delete the monitoring from the CloudWatch dashboard.