A company recently migrated its application to an Amazon Elastic Kubernetes Service (Amazon EKS) cluster that uses Amazon EC2 instances. The company configured the application to automatically scale…

Attaching the `CloudWatchAgentServerPolicy` to the IAM instance profile of the EKS EC2 nodes grants the necessary permissions for the unified CloudWatch agent to publish collected metrics and logs to CloudWatch.

While service accounts are used for pods, the CloudWatch agent for Container Insights typically runs with permissions inherited from the underlying EC2 instance profile, making the instance profile the more direct and common place to attach the policy for node-level agents.

Deploying the unified Amazon CloudWatch agent to the EKS cluster is the standard and recommended method for collecting detailed performance metrics, including memory usage, from Kubernetes components and application pods for Container Insights.

While AWS Distro for OpenTelemetry (ADOT) can collect metrics, the unified Amazon CloudWatch agent (Option C) is the native solution for feeding metrics directly into CloudWatch Container Insights, which is the specified analysis platform.

Analyzing the `pod_memory_utilization` metric within the CloudWatch Container Insights dashboard directly addresses the application's memory errors by providing insight into resource consumption at the individual pod level, which is critical for identifying bottlenecks and adjusting scaling.

While `node_memory_utilization` is important for node health, `pod_memory_utilization` (Option E) is more directly relevant to diagnosing memory errors within a specific application running in pods and informing application-level scaling decisions.