Realtime Compute for Apache Flink:January 9, 2026

SQL enhancements

Bitmap type support — The new native Bitmap type and its initial set of functions (bitmap construction and cardinality statistics) enable efficient, real-time, precise deduplication—for example, unique visitor (UV) counting.

Data ingestion via YAML (Flink CDC)

• Metadata reference — Directly reference source and sink tables registered in the Catalog within Flink CDC YAML scripts, without redefining table schemas inline.

• Complex routing logic — Configure table names in Route using standard regular expressions to support sharded table merging scenarios.

• Dirty data handling — JSON format parsing now supports a dirty data handling mechanism, so malformed records no longer cause job failures.

• Automatic precision expansion — StarRocks now partially supports automatic precision expansion for the Decimal type, reducing manual schema adjustments for high-precision data.

• Empty schema fault tolerance — The Kafka source can now create tables with an empty schema when the source has no data yet, preventing job startup failures in early-stage pipelines.

• Paimon source support — Apache Paimon is now officially supported as a Flink CDC source, enabling change data capture directly from Paimon tables.

Connectors

• MySQL — Three enhancements in this release:

  • Throttling: A data throttling feature prevents excessive load on the source database during high-throughput ingestion.

  • Metadata extension: The file and pos binary logging metadata fields are now supported. Use these fields to determine the modification order of the same row across events.

  • Performance optimization: In snapshot-only mode, the connector no longer executes binary logging-related commands, reducing overhead when binary logging is not needed.

• PolarDB-X CDC — Public preview. Two enhancements:

  • Subscribe to binary logging streams with multiple concurrent connections to significantly increase ingestion throughput.

  • Subscribe to binary logging at the table level for finer-grained CDC control.

• OceanBase — Public preview. The new bypass import write feature significantly improves write performance for large data batches.

• MongoDB — The sink table now supports Partial Update, so only changed fields are written rather than full documents.

• Elasticsearch — Two enhancements:

  • Explicitly configure the doc_as_upsert parameter to control upsert behavior.

  • Configure a connection timeout parameter to handle unstable network environments.

• Apache Kafka — Redundant consumer groups created by the Kafka connector are now automatically deleted, reducing clutter in your Kafka cluster.

Data lakehouse integration (Iceberg)

Iceberg Connector — Configurable Hadoop-related parameters give you greater flexibility when connecting to complex Hadoop environments.