DEVOURED
Amazon CloudWatch launches OTel Container Insights for Amazon EKS
Amazon CloudWatch adds OpenTelemetry-based monitoring for EKS clusters with enriched high-cardinality metrics and PromQL query support.
What: CloudWatch Container Insights now supports OpenTelemetry metrics for Amazon EKS clusters in public preview, automatically enriching each metric with up to 150 labels including Kubernetes metadata and custom tags, with curated dashboards and PromQL querying capabilities.
Why it matters: This shifts AWS monitoring toward open standards (OpenTelemetry and PromQL) rather than proprietary formats, and high-cardinality metrics enable more granular filtering and analysis by custom dimensions like team, application, or infrastructure attributes.
Takeaway: EKS users can install the CloudWatch Observability add-on via one-click in the EKS console or infrastructure-as-code tools to get enhanced monitoring free during preview in five AWS regions.
Deep dive
- CloudWatch Container Insights now collects metrics using the OpenTelemetry Protocol (OTLP), embracing open observability standards instead of AWS-specific collection methods
- Each metric is automatically enriched with up to 150 descriptive labels, including Kubernetes metadata (namespace, pod name, etc.) and custom labels developers can define for team, application, or business unit categorization
- Curated dashboards provide pre-built visualizations for cluster, node, and pod health with filtering capabilities by instance type, availability zone, node group, or any custom label applied to resources
- CloudWatch Query Studio now supports PromQL (Prometheus Query Language) for writing custom queries, making it familiar for teams already using Prometheus-based monitoring
- The CloudWatch Observability EKS add-on enables one-click deployment through the EKS console, or can be provisioned via CloudFormation, CDK, or Terraform for infrastructure-as-code workflows
- Hardware auto-detection automatically identifies and monitors accelerated compute resources including NVIDIA GPUs, Elastic Fabric Adapters, and AWS-specific AI chips (Trainium and Inferentia)
- Existing Container Insights users can run both OpenTelemetry and legacy Container Insights metrics in parallel during migration, avoiding breaking changes to existing monitoring
- Available in five regions during preview: US East (N. Virginia), US West (Oregon), Asia Pacific (Sydney and Singapore), and Europe (Ireland)
- No charges apply for OpenTelemetry metrics from Container Insights during the preview period, making it risk-free to test before general availability pricing is announced
Decoder
- EKS: Amazon Elastic Kubernetes Service, AWS's managed Kubernetes offering that handles control plane operations
- OpenTelemetry (OTel): Open-source observability framework for collecting and exporting metrics, traces, and logs in a vendor-neutral format
- OTLP: OpenTelemetry Protocol, the standardized format for transmitting observability data from collectors to backends
- High-cardinality metrics: Metrics with many unique label combinations (like per-pod or per-user metrics), which traditional monitoring systems struggle to handle at scale
- PromQL: Prometheus Query Language, the widely-used query language for filtering and aggregating time-series metrics
- Container Insights: AWS's monitoring solution specifically designed for containerized applications on ECS, EKS, and Fargate
- Trainium/Inferentia: AWS-designed machine learning accelerator chips for training and inference workloads
Original article
Amazon CloudWatch adds Container Insights with OpenTelemetry metrics for Amazon EKS, delivering enriched, high-cardinality metrics and flexible querying via PromQL with curated dashboards. It supports easy deployment, hardware auto-detection, and dual metric publishing, offering enhanced observability at no cost during preview.
Explore all 55 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi