victoria-metrics-cluster

VictoriaMetrics cluster (vminsert + vmselect + vmstorage)

VictoriaMetrics cluster mode — the horizontally scalable variant of the VictoriaMetrics time-series database. The cluster decomposes the single-node binary into three independent services that communicate over RPC and follow a shared-nothing architecture: - vmstorage stores raw samples and serves queries filtered by time range and labels. Each vmstorage node is autonomous; nodes do not communicate with each other. - vminsert receives ingested data and shards it across vmstorage nodes via consistent hashing on metric name and labels. - vmselect runs queries by fanning out to all configured vmstorage nodes and merging the results. Each service can be scaled independently. Cluster mode also adds multi-tenancy (accountID/projectID namespaces in URLs and labels) and optional replication via -replicationFactor. Built from source using the Gentoo Go toolchain. For ingestion rates below ~1M samples/sec the single-node sibling (app-metrics/victoria- metrics) is recommended — it is simpler to operate and has lower per-node overhead.

homepage ↗homepage ↗