Choose managed AWS-native tooling for speed, or OpenTelemetry for portability.

Container Insights

Fastest path to CloudWatch dashboards for nodes, Pods and clusters. Install the CloudWatch Observability add-on.

ADOT / OpenTelemetry

Vendor-neutral traces and metrics you can send to CloudWatch, X-Ray, Prometheus or a third party without re-instrumenting.

Fluent Bit

The light log shipper (C, small memory footprint) on every node, to CloudWatch Logs, S3 or another destination.

Managed Prometheus and Grafana

Standard PromQL and dashboards without running Prometheus storage yourself. Good fit for SLOs.

Observability on EKS: metrics, logs and traces

Observability has three signals and EKS gives several ways to collect them. Container Insights through the CloudWatch Observability add-on is the quickest path to dashboards and alarms if CloudWatch is your standard backend. The AWS Distro for OpenTelemetry (ADOT) collector gives vendor-neutral metrics and traces that you can route to CloudWatch, X-Ray, Amazon Managed Service for Prometheus or a third-party platform. Fluent Bit, a lightweight log processor, runs as a DaemonSet and ships container logs.

Amazon Managed Service for Prometheus and Amazon Managed Grafana remove the burden of running Prometheus storage and Grafana. They are a good fit for teams that already use Prometheus conventions and want SLO dashboards and alerts. Reveal the four options with the right arrow and see how they connect to the same Pods.

Choose by what you value. If speed and tight AWS integration matter, use Container Insights. If portability matters, instrument with OpenTelemetry and treat the backend as replaceable. In either case instrument the golden signals (latency, traffic, errors, saturation) per service and alert on SLO burn rates, not on every CPU spike. Deck 6 builds on this with capacity and cost visibility.

Cost note: logs and metrics can become a major bill. Control cardinality, set retention, sample traces, and drop noisy debug logs at the collector.