victoriametrics-unused-metrics-analysis

You can find unused VictoriaMetrics metrics by querying the metric_names_stats API for series with zero query requests and old lastRequestTimestamp values, then dropping them via vmagent relabel_configs to reduce storage and ingestion costs.

Metric cardinality optimization identifies and removes rarely queried or unused time-series in Prometheus-compatible systems to reduce wasted ingestion. You need it when storage costs rise or scrape targets generate excessive unqueried metric data.

Group unused metrics by prefix using metric_names_stats output, then apply vmagent relabel_configs or the -relabelConfig flag to drop matching metric names. This ensures only actively queried series are ingested going forward.

Yes, the metric_names_stats API targets standard Prometheus-compatible metric names. It analyzes usage across prefixes and filters to truly unused metrics, supporting workflow scenarios including dashboards, alerts, and recording rules.

Dropping unused metrics requires caution because series with zero queryRequests might still have dependencies with dashboards, alerts, or recording rules. Always verify that truly unused metrics have no downstream workflow dependencies before removal.