All Products
Search

This Product

    Realtime Compute for Apache Flink:Version 2025-10-13

    Document Center

    Realtime Compute for Apache Flink:Version 2025-10-13

    Last Updated:Mar 26, 2026

    This topic describes the major feature updates and bug fixes in the Realtime Compute for Apache Flink release on October 13, 2025.

    Important

    The version upgrade is gradually released to users. For more information, see the latest announcement in the Realtime Compute for Apache Flink console. You can use the new features in this version only after the upgrade is complete for your account. To apply for an expedited upgrade, submit a ticket.

    Overview

    This release delivers engine and platform updates, including new connector capabilities, AI vector search, SQL improvements, performance optimizations, and bug fixes.

    Engine

    Data ingestion

    Kafka source: The Canal-JSON format now parses native MySQL_TYPE fields, improving type fidelity when consuming MySQL change events.

    Connectors

    • Hologres connector: You can now specify the type of binary log to read for more granular data sync control.

    • StarRocks connector: Updated to align with StarRocks Connector for Flink release 1.2.11.

    • Lindorm connector: Dimension tables now support one-to-many joins.

    • MaxCompute connector: You can now set a custom time zone for cross-time-zone data processing.

    • SLS connector: The processor parameter lets you pre-filter Simple Log Service (SLS) data before consumption, reducing downstream processing costs.

    AI capabilities

    • Vector search: The VECTOR_SEARCH function runs real-time similarity searches using Milvus. See VECTOR_SEARCH.

    • AI function: More configurable parameters are available for greater control over AI function behavior.

    SQL and data management

    • JSON conversion: Use the JSON_QUERY function to convert an array in a JSON string directly to ARRAY<STRING>. See JSON functions.

    • StarRocks Catalog: You no longer need to explicitly pass a URL when using StarRocks Catalog.

    Performance optimizations

    • MySQL connector: Predicate pushdown for boolean types is optimized, reducing unnecessary data transfer.

    • Hologres connector: Projection pushdown and data compression are now applied when consuming binary logs, reducing network and processing overhead.

    • MaxCompute connector: Write performance for complex types such as arrays into regular tables is improved.

    Platform

    New capabilities

    Capability Status Region
    Workflows Public preview US (Silicon Valley)
    Hybrid billing for YITIAN ARM workspaces Now available China (Ulanqab)

    Platform experience enhancements

    Purchase experience

    The 1,000 compute unit (CU) limit for subscription workspaces has been removed. This applies to scale-ups, new purchases, and conversions from pay-as-you-go to subscription, enabling elastic scaling for large-scale jobs.

    Observability

    • Unified garbage collection (GC) metric display: GC metrics now support ParallelGC and CMS, enabling unified monitoring across GC types.

    • New busyTimePerSecond metric: Provides precise insights into job load to help you identify performance bottlenecks, allocate resources effectively, and ensure job stability.

    Permission model

    Folder permissions are now unified with job draft and data query permissions. You can create folders without encountering permission errors.

    Materialized table

    When editing a materialized table, you can now upgrade the Ververica Runtime (VVR) version inline. The editor also supports version comparison and automatically updates the job deployment.

    Message delivery configuration

    You can customize the delivery scope for messages to achieve more granular control and reduce notification costs.

    Feature summary

    Feature Description Reference
    Hologres concurrent batch writes Concurrent writes to Hologres tables with a primary key after a reshuffle, increasing write throughput. Connector options (VVR 11 or later)
    Hologres binary log consumption optimization Projection pushdown and data compression when consuming binary logs, reducing network and processing overhead. —
    Specify Hologres binary log type Specify the type of binary log to read for more granular data sync. —
    Kafka Canal-JSON MySQL_TYPE parsing The Canal-JSON format for the Kafka source now parses native MySQL types. Kafka
    JSON String to ARRAY<STRING> conversion Use JSON_QUERY to convert an array in a JSON string to ARRAY<STRING>. JSON functions
    AI vector search The VECTOR_SEARCH function runs real-time similarity searches using Milvus. VECTOR_SEARCH
    MaxCompute time zone setting Custom time zones for cross-time-zone data processing. MaxCompute
    SLS consumption processor The processor parameter pre-filters SLS data before consumption, reducing downstream processing costs. SLS

    Notable fixes

    Connector fixes

    • PostgreSQL CDC: Fixed a deadlock when reading data with PostgreSQL CDC.

    • Kafka connector: Fixed an issue where the connector dropped time zone information when converting timestamps with time zone attributes.

    • Lindorm connector: Resolved a NullPointerException during joins with a Lindorm dimension table.

    • Paimon connector: Fixed an error where viewing or deleting a Paimon table after a parameter misconfiguration returned Could not find any factory for identifier 'last_not_null_value'.

    SQL fixes

    • JSON_QUERY function: Fixed a format issue with the RETURNING ARRAY<STRING> clause.