LoginSign Up
← Back to all sparks
A

Apache SeaTunnel

ANALYTICS
Velocity0.0

High-performance distributed data integration platform

seatunnel.apache.org ↗

SeaTunnel can finally split one large file across readers — and hasn't shipped since March.

data integrationparallel readscdcconnectorsapache projectslowing cadence
◆Current state
The 2.3.13 release in March is by far the densest in this window: parallel splitting of large files for HDFS, local CSV/text/JSON and logical Parquet splits, CDC source schema evolution on the Flink engine, a checkpoint API with configurable minimum pause, and new connectors for DuckDB, Lance, AWS DSQL and HugeGraph. The releases before it were thinner — 2.3.12 and 2.3.11 are dominated by documentation, much of it Chinese translations of existing connector pages, and 2.3.9 and 2.3.8 are bug fix rollups.
◆Where it's heading
Two things are happening at once. The engine is getting faster on the shapes that actually stall a pipeline — a single enormous file, a schema that changed under a running CDC job — and the connector catalogue keeps widening toward analytical and vector-adjacent stores rather than more transactional databases. But the cadence has stretched: releases used to land every two to three months, and nothing has shipped in nearly five.
◆Prediction
Expect the split-and-parallel-read work started for files to extend to more source connectors, since it is the change with the broadest effect on throughput. The release gap is the open question — these entries show a lengthening interval without indicating whether a 2.4 line is being prepared behind it.

◆Recent moves

  1. 4mo ago

    Large files split for parallel read; CDC schema evolution on Flink

    ⚡ SPARK

    The release that lifts SeaTunnel past a structural limit — one file no longer means one reader — while adding schema evolution for CDC sources on Flink. It is also the last release the project has published, which makes it both the high point and the start of the current quiet stretch.

    View source ↗
  2. 10mo ago

    Documentation sweep with JDBC and Iceberg updates

    A release weighted toward documentation — broken links, DynamoDB parameter corrections, a Zeta tuning guide, and Iceberg docs updated for S3 Tables REST catalogues. Useful, but it is the connector docs catching up with connectors that already shipped.

    View source ↗
  3. 1y ago

    Chinese connector documentation added in bulk

    Almost entirely Chinese translations of existing connector documentation — MySQL CDC, MongoDB CDC, HiveJdbc, Jira and others. It reflects where the contributor base is, and changes nothing functional.

    View source ↗
  4. 1y ago

    Connector option handling normalized across the catalogue

    A sweep through connector configuration — InfluxDB, HTTP, IoTDB, email, Web3j and local file options all reworked, with one change reverted in the same release. Tidying config surfaces across dozens of connectors is unglamorous and exactly what a catalogue this size needs.

    View source ↗
  5. 1y ago

    Type conversion and connector bug fixes

    A bug fix rollup: Avro short-to-int conversion, a null pointer when source fields are fewer than Hive target fields, MongoDB BSON conversion and a Hadoop Kerberos failure. Individually narrow, collectively the long tail of a wide connector surface.

    View source ↗
  6. 1y ago

    Multi-table sink concurrency and Paimon fixes

    Core concurrency exceptions in multi-table sink commit preparation, a Windows script syntax error, and Paimon dynamic bucket handling. Maintenance across engine and connectors with nothing new attached.

    View source ↗