Metrics

Definition

Metrics are numeric measurements that track the state or behaviour of a system over time. They are stored as time series — a sequence of (timestamp, value) pairs — and are one of the three core telemetry signals alongside logs and traces. Common metric types include counters, gauges, and histograms.

In depth

Metrics are the most efficient signal type for long-term trending and alerting at scale: because they are aggregated numbers rather than individual events, a single metric can represent millions of requests with minimal storage. Counter metrics track cumulative totals (total HTTP requests), gauges track point-in-time values (current memory usage), and histograms track value distributions (request latency buckets). The four golden signals — latency, traffic, errors, and saturation — are all expressible as metrics. Metrics pair naturally with traces and logs: a metric alert fires, an SRE investigates the correlated traces, and finds the root cause in a log line. obseria.io stores metrics in a columnar TSDB with 13-month default retention and full PromQL support.

Related terms