InfluxDB is a time-series database purpose-built for metrics, events, and other timestamped data. Data is written using the InfluxDB line protocol — a simple text format of measurement,tag=value field=value timestamp — typically via Telegraf, the project’s companion collection agent that ships with hundreds of input plugins.
InfluxDB 1.x used a custom TSM storage engine and the InfluxQL query language. 2.x added Flux, a functional data-scripting language, plus a tasks/alerting UI. InfluxDB 3 is a complete rewrite in Rust on top of Apache Arrow, DataFusion, and Parquet: data is held in-memory in Arrow format, persisted as Parquet files to object storage, and queried with SQL via DataFusion (with InfluxQL still supported). The Core edition is open source under MIT/Apache-2.0 dual license; clustering, replication, and long-term historical query live in the commercial Enterprise and Cloud editions.
In the observability stack it is typically paired with Grafana for dashboards and Telegraf or an OpenTelemetry collector for ingestion. Compared to Prometheus it is a true pushed-ingest database with much higher cardinality tolerance in 3.x, and compared to TimescaleDB it is not a Postgres extension — you get a dedicated engine but give up the rest of the Postgres ecosystem.