metrics-design

Prometheus, OpenTelemetry, Datadog, and CloudWatch are all supported for metrics design and instrumentation. The Skill provides specific guidance on defining naming conventions, low-cardinality labels, and recording rules across these monitoring stacks.

Choosing the right metric types like Counters, Gauges, Histograms, and Summaries depends on the operational signal you need to track. Counters measure cumulative events, Gauges capture instantaneous values, and Histograms or Summaries record distributions such as request durations.

Designing a metrics plan for a new service involves selecting appropriate metric types, defining naming conventions, and identifying RED/USE signals. After finalizing the design, you implement the instrumentation directly in your codebase to capture production observability data.

Low-cardinality labels are required for Prometheus recording rules and dashboards to prevent unbounded time series growth. High cardinality degrades query performance and increases storage costs, making strict label schema control essential for production metrics.