// the find
influxdata/influxdb
Scalable datastore for metrics, events, and real-time analytics
InfluxDB 3 Core is a time series database rebuilt from scratch in Rust, using Apache Arrow, DataFusion, and Parquet as its storage and query foundation. It's aimed at teams collecting high-frequency metrics, IoT sensor data, or any event stream where you need fast last-value and range queries. This is a significant architectural departure from v1/v2 — if you're on those versions, migration is not trivial.
The Arrow/DataFusion/Parquet stack is a genuinely good call: you get columnar storage, vectorized query execution, and interoperability with the broader Arrow ecosystem for free. Diskless architecture with object storage (S3/GCS/Azure) as the primary persistence layer means you can scale storage independently of compute and stop worrying about disk provisioning. The embedded Python VM for triggers and plugins is practical — you can do light transformations at ingest without a separate stream processor. Backward compatibility with v1/v2 write APIs means existing Telegraf/client pipelines work without changes.
The version fragmentation is a real operational headache: v1 on one branch, v2 on another, v3 here, and they are not drop-in compatible on the query side — InfluxQL works but Flux (the v2 query language) is absent from v3, so any non-trivial v2 query logic needs to be rewritten in SQL. No multi-node clustering in Core; that's Enterprise-only, so if you outgrow a single node you're either paying or rearchitecting. The processing engine (Python plugins) is still rough — the README points to a separate markdown file for it, which usually means 'works but not polished.' Build times from source are brutal for a Rust monorepo this size, and the CI is CircleCI-only with no obvious path for contributors using GitHub Actions.