Victoria Metrics

VictoriaMetrics is the most resource-efficient open-source time-series database — uses 3-10x less RAM and 2-4x less disk than Prometheus for equivalent data, handles higher cardinality, and provides better query performance at scale. It's Prometheus-compatible: Grafana connects to VictoriaMetrics as a Prometheus data source, and all PromQL queries work. MetricsQL extends PromQL with additional aggregation functions. For agents: ingest metrics via Prometheus remote_write (no code change from Prometheus-instrumented applications), query via HTTP API using PromQL/MetricsQL, and benefit from long-term storage without Thanos complexity. Self-hostable as a single binary. Confidence is docs-derived.

HTTP API at {victoria_host}. Metrics ingestion: POST /api/v1/import/prometheus (Prometheus text format lines). POST /api/v1/import (JSON format). Prometheus remote_write target: configure in prometheus.yml or any remote_write-compatible agent. InfluxDB line protocol: POST /write. OpenTelemetry: POST /opentelemetry/v1/metrics (OTLP). Querying: GET /api/v1/query?query={metricsql_expr}&time={timestamp}. GET /api/v1/query_range?query={expr}&start={}&end={}&step={}. GET /api/v1/series?match[]={selector} discovers series. GET /api/v1/label/{name}/values retrieves label values. Prometheus-compatible API: Grafana connects natively as Prometheus data source.

Default: no authentication (same as Prometheus — add at ingress/reverse proxy level). Basic auth + TLS configurable via flags (-httpAuth.username, -httpAuth.password). Bearer token: pass via Authorization header (if configured). Enterprise version: LDAP/AD auth available. Recommendation: deploy behind authenticated reverse proxy for any exposed instance. Cluster version: additional auth layer for inter-component communication. Self-hosted: operator-responsible for auth and network security.

Standard HTTP status codes with JSON responses (Prometheus-compatible). PromQL errors: syntax error with position info. High cardinality handling: VictoriaMetrics handles higher cardinality than Prometheus natively. Ingestion backpressure: 429 if ingestion rate exceeds capacity — tune retention and storage. Single-node: automatic WAL recovery on restart. Cluster version: higher availability with multiple replicas. VictoriaMetrics Cloud uptime (managed): status.victoriametrics.com. Self-hosted: operator-managed. Memory usage: significantly lower than Prometheus for equivalent cardinality — major operational advantage.

docs.victoriametrics.com is comprehensive — single-node quickstart, cluster configuration, MetricsQL reference, Prometheus migration guide, API reference, and Grafana integration. Getting started: docker run victoriametrics/victoria-metrics — running in 30 seconds; Prometheus remote_write configured in 5 minutes. Single binary with no external dependencies. Community via VictoriaMetrics GitHub (12k+ stars, very active) and Slack. MetricsQL documentation explains extensions over PromQL. Grafana dashboards for VictoriaMetrics self-monitoring available on Grafana.com.