All Products
Search
Document Center

Realtime Compute for Apache Flink:December 12, 2022 release

Last Updated:Mar 26, 2026

This release is based on Ververica Runtime (VVR) 6.0.3 of Apache Flink 1.15.2. It fixes major defects from the Apache Flink community and includes platform improvements and connector updates.

Important

If your deployment runs a VVR version earlier than 6.0.3, upgrade to VVR 6.0.3 as soon as possible to ensure stable production.

New features

Platform

Query errors directly from log selections

On the Events and Exploration tabs in the Realtime Compute for Apache Flink development console, select text in a log entry to query the related error. The results show the cause and recommended solution, so you can troubleshoot without leaving the console.

Optimize SQL syntax error messages

SQL syntax verification now returns clearer error messages, making it faster to identify and fix invalid syntax.

Connectors and catalogs

Configure a retry mechanism for MaxCompute writes

Set a retry policy for data writing to MaxCompute to prevent failures of the COMMIT or createUploadSession operation when service interruptions occur in a Tunnel cluster. This keeps your write jobs resilient without manual intervention.

Create or delete non-nullable columns in `CREATE DATABASE AS` deployments

Deployments that use the CREATE DATABASE AS statement now support creating and deleting non-nullable columns.

Align Kafka JSON catalog schema with JSON message column order

The schema inferred by a Kafka JSON catalog now matches the column sequence in the source JSON message, eliminating schema inconsistencies caused by column reordering.

Tolerate invalid table creation syntax in ApsaraDB RDS

Realtime Compute for Apache Flink is now compatible with invalid table creation syntax in ApsaraDB RDS databases. Running Change Data Capture (CDC) tasks are not interrupted when invalid syntax is encountered.

Bug fixes

`sink.parallelism` setting ignored in `CREATE DATABASE AS` to Kafka sync

The sink.parallelism parameter now takes effect when you use CREATE DATABASE AS to synchronize data to Kafka.

`ClassCastException` in `CREATE DATABASE AS` deployments (VVR 6.0.2)

Deployments using CREATE DATABASE AS no longer throw a ClassCastException error.

Latency metric charts not displayed for MySQL CDC source tables (VVR 6.0.2)

Curve charts for latency-related metrics of MySQL CDC source tables now render correctly.

SQL hints syntax not applied (VVR 6.0.2)

SQL hints now take effect as expected during query execution.

Class conflict reported as data out-of-range exception during syntax checks (VVR 6.0.2)

Class conflicts during syntax checks are now reported accurately instead of being misclassified as data out-of-range exceptions.

Server ID conflict when writing from one table to multiple sinks via `CREATE TABLE AS` (VVR 6.0.2)

Server IDs no longer conflict when a CREATE TABLE AS deployment writes data from one table to multiple sinks.

JobManager deadlock due to leader election in Kubernetes clusters

The JobManager no longer deadlocks during leader election in Kubernetes clusters.

"Decrease reference to an unknown file" error in GeminiStateBackend

The GeminiStateBackend state backend no longer produces "Decrease reference to an unknown file" errors.