Telemetry correlation links related observability signals so users can move from one record or view to corresponding evidence in another signal type. Frontends commonly provide the user-facing operations and visualizations for these transitions. section 6.4 section 6.3
Common correlation paths include:
- Metrics to traces: use a metric from a dashboard or alert to find representative spans from contributing services. section 11.1.3
- Metrics to metrics: move between related measurements, such as client-side and server-side errors. section 11.1.3
- Metrics to logs: move from an aggregate metric value to log lines produced by contributing services. section 11.1.3
- Traces or metrics to profiles: connect distributed traces or metrics to label-based profiles or frontend-supported profile views. section 9.3.4
A practical metric-to-trace path
A dashboard metric or alert can provide the starting point, while representative spans from services contributing to that time series provide the destination. Prometheus, Jaeger, and Grafana can implement this as exemplar-based correlation. section 11.1.3
Correlation is useful but unevenly implemented. Check which paths a frontend supports, whether signals use compatible formats and terminology, and how much manual work separates one signal type from another. Metadata such as Prometheus labels, Kubernetes labels, and OpenTelemetry resource or signal attributes supplies the connective tissue; OpenTelemetry-based systems can improve interoperability and reduce vendor lock-in. section 6.3 section 11.4.1
