CloudWatch Container Insights Service | Overview.

Amazon CloudWatch Container Insights Service - Overview.

Scope:

• Intro,
• Key Features & Capabilities,
• Enabling Container Insights,
• Links to official documentation,
• The concept: CloudWatch Container Insights,
• How Container Insights Works,
• Data Collection,
• Metrics Processing,
• Storage & Visualization,
• Alarms & Automation,
• Data Collected Samples,
• Security & Permissions,
• Advanced Use Cases,
• Architecture Diagram,
• Final thoughts.

Intro:

• Amazon CloudWatch Container Insights is a monitoring and observability service that:
•  Collects, 
• Aggregates, 
• Summarizes metrics / logs
•  From containerized applications and microservices. 
• Amazon CloudWatch Container Insights Service provides visibility into the performance and health of container management services, including: 
• Amazon Elastic Container Service (Amazon ECS)
• Amazon Elastic Kubernetes Service (Amazon EKS)
• AWS Fargate
• Standalone Kubernetes.

Key Features and Capabilities

• Automatic Dashboards: Container Insights provides pre-built, automatic dashboards that display key metrics such as CPU utilization, memory usage, network data, and disk I/O at various levels (cluster, node, pod, task, and service).
• Troubleshooting and Diagnostics: It helps in isolating issues quickly by providing diagnostic information, such as container restart failures.
• Enhanced Observability: Recent updates offer enhanced observability features, including new landing pages and the ability to view resource consumption sorted by different parameters, aiding in proactive risk mitigation.
• Log and Metric Collection: Operational data is collected as performance log events using the embedded metric format, from which CloudWatch then generates aggregated metrics.
• Integration with OpenTelemetry: AWS Distro for OpenTelemetry (ADOT) can be used to collect system metrics, providing a secure and supported distribution of APIs and agents for unified observability.
• Alarms and Notifications: Users can set CloudWatch alarms on the collected metrics to notify teams of performance issues or resource constraints. 

Enabling Container Insights

• twtech can enable Container Insights through the AWS console, AWS CLI, CloudFormation, or CDK. 
• EKS and Kubernetes Container Insights service, is often  set up quickly using the Amazon CloudWatch Observability EKS add-on. 

https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/Container-Insights-setup-EKS-quickstart.html

Link to Container Insights official documentation: Amazon CloudWatch Container Insights documentation.

https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/ContainerInsights.html

The concept: CloudWatch Container Insights

• CloudWatch Container Insights is an observability feature that automatically collects, aggregates, then summarizes metrics / logs from containerized applications and microservices which are running on:
• Amazon ECS (EC2 & Fargate)
• Amazon EKS
• Kubernetes (self-managed on EC2)
• AWS App Runner.

NB:

• CloudWatch Container Insights Service helps twtech to monitor, troubleshoot, and optimize container workloads without building custom telemetry pipelines.

 How Container Insights Works

1. Data Collection

• Uses CloudWatch Agent (with embedded Fluent Bit) deployed as:
• DaemonSet on EKS/Kubernetes
• Sidecar or embedded agent in ECS/ECS on Fargate
• Collects:
• Performance Metrics: CPU, memory, disk, network per container/pod/service/node/cluster
• Logs: stdout/stderr logs, kube-system logs
• Diagnostic Data: container restart counts, OOM events, throttling

2. Metrics Processing

• Metrics are aggregated at multiple levels:
• Cluster
• Node/EC2 instance
• Pod/Task
• Container
• Service/Namespace
• Automatically published into CloudWatch Metrics namespace:
  ContainerInsights/ClusterName

3. Storage & Visualization

• Metrics → CloudWatch Metrics
• Logs → CloudWatch Logs
• Logs can be queried using CloudWatch Logs Insights
• Dashboards auto-created in CloudWatch for:
• Cluster overview
• Nodes
• Pods/Tasks
• Services

4. Alarms & Automation

• Create CloudWatch Alarms on container metrics (e.g., CPU > 80%, OOMKills > 0).
• Integrates with EventBridge for automated responses (scaling, healing, notifications).

Data Collected Samples

Level	Metrics Examples
Cluster		CPU reservation %, memory reservation %, pod count, service count
Node		CPU utilization, memory utilization, disk I/O, network traffic, running pods
Pod/Task		CPU/memory limits vs requests, restarts, OOM events
Container		CPU usage (cores, %), memory (RSS, working set), network bytes in/out
Service		Aggregated CPU/memory usage across replicas

 Security & Permissions

• IAM permissions required for:
• Writing to CloudWatch Logs
• Publishing metrics to CloudWatch
• Best practice: use least privilege IAM roles for service accounts (IRSA in EKS).

 Advanced Use Cases

1. Performance Troubleshooting
• Find “noisy neighbors” in multi-tenant clusters
• Detect throttled pods (CPU throttling metrics)
• Spot OOM-killed containers
2. Cost Optimization
• Identify underutilized nodes or over-provisioned requests/limits
• Right-size pods and tasks to reduce EC2/Fargate costs
3. Scaling Decisions
• Feed Container Insights metrics into Application Auto Scaling
• Custom autoscaling policies (beyond CPU/memory)
4. Compliance & Auditing
• Centralized logs across all namespaces/clusters
• Integrate with CloudWatch Contributor Insights for anomaly detection

 Architecture Diagram 

[Cluster (EKS/ECS)]
 CloudWatch Agent/Fluent Bit DaemonSet (Collects metrics & log then Sends Metrics) → CloudWatch Metrics (Sends Logs) → CloudWatch Dashboards, Alarms, Logs Insights → EventBridge, Lambda, Auto Scaling (Integration).

Final thoughts:

• CloudWatch Container Insights Service is the observability out-of-the-box for containerized workloads.
• CloudWatch Container Insights Service has minimal setup, tightly integrated with AWS monitoring and automation.