InfluxDB 3 Enterprise release notes

v0.1.0 Beta

Core

Features

• Query and storage enhancements

  • New ability to stream response data for CSV and JSON queries, similar to how JSONL streaming works.
  • Parquet files are now cached on the query path, improving performance.
  • Query buffer is incrementally cleared when snapshotting, lowering memory spikes.

• Processing engine improvements

  • New scheduled trigger enables Python plugins to run on a custom, time-defined basis.
  • New request trigger enables calling Python plugins via HTTP requests.
  • New in-memory cache for storing data temporarily; cached data can be stored for a single trigger or across all triggers.
  • Integration with virtual environments and install packages. You can now specify a Python virtual environment via the CLI or VIRTUAL_ENV variable and install packages or a requirements.txt.
  • Python plugins are now implemented through triggers only. Simply create a trigger that references your Python plugin code file directly.
  • Snapshots are now persisted in parallel, improving performance by running jobs simultaneously, rather than sequentially.
  • Write to logs from within the Processing Engine.

• DB and CLI improvements

  • You can now specify the precision on your timestamps for writes using the --precision flag. Includes nano/micro/milli/seconds (ns/us/ms/s).
  • Added a new show system subcommand to display system tables with different options via SQL (default limit: 100)
  • Clearer table creation error messages

Bug fixes

• If a database was created and the service was killed before any data was written, the database would not be retained.
• A last cache with specific “value” columns could not be queried.
• Running CTRL-C no longer stopped an InfluxDB process, due to a Python trigger.
• A previous build had broken JSON queries for RecordBatches.
• There was an issue with the distinct cache that caused panics.

Enterprise

Features

• Cluster management
  • Nodes are now associated with clusters, simplifying compaction, read replication, and rocessing.
  • Node specs are now available for simpler management of cache creations.
• Mode types
  • Ingest, query, compaction, and processing engine can now all be set individually per node.