A metric is a numerical value used to describe service behavior over time. In Prometheus, a metric has a name, value, label, and timestamp. Timestamps reveal changes; labels let you group measurements. section 9.1.2
Metrics help make a distributed application visible. Service-level measurements and overall latency can expose trouble even when a request succeeds. Use dashboards to inspect trends, then add alerts for conditions that need attention when nobody is watching. section 1.4.4
From measurement to action
Prometheus can collect system and application metrics. An alert might notify the team when a service’s memory usage exceeds 80%. For customer-facing quality, request latency, error rate, and throughput can serve as service-level indicators. Check how latency is summarized: an average can obscure a small number of very slow responses. section 1.4.4 section 9.1.2
Metrics are one part of observability, alongside traces and logs. When a metric shows an anomaly, related traces can identify the request flow and logs can provide context from the same period. OpenTelemetry instrumentation can ship metrics to a collector for monitoring. section 9.4 section 9.0
