Metrics

Use Metrics for rates, gauges, distributions, and trends that should be cheaper to read than raw event streams.

Query metric points

Open Metrics or Explore and start with the metrics source.

Examples:

metrics | where name == "http.server.duration" | stats p95(value) by service, bin(auto)

metrics | where name == "process.runtime.go.goroutines" | stats max(value) by service

metrics | where temporality == "cumulative" | fields time, name, service, value, unit

Metric rows expose instrument context such as point_kind, temporality, is_monotonic, unit, scope_name, and scope_version.

Permissions

Metrics require telemetry:query.

Pivots

• Use exemplars to pivot from a metric point to a trace when exemplar data is present.
• Pin stable metric queries to dashboards.
• Use service filters to compare a metric across environments.
• Convert threshold-style metric queries into alert rules.