Realtime Compute for Apache Flink:August 8, 2025

Engine

Ververica Runtime (VVR) 11.2 is now available. Built on Apache Flink 1.20.2, VVR 11.2 provides additional optimizations and enterprise enhancements. Highlights:

AI function

Introduces configurable strategies for handling messages that exceed the AI model's maximum context window: Discard or Truncate.

Connector

• MySQL CDC connector: Optimizes handling of VARCHAR fields, improving synchronization performance and stability.

• Now supports the Canal-JSON format for data ingestion. Can extract both event timestamps (ts) and event sequences (es) fields.

• AnalyticDB for MySQL connector: Adds support for INSERT IGNORE, improving fault tolerance during data writes.

Security

The Paimon and OSS connectors now support RAM-based authorization, replacing AccessKey pairs for improved security and permission management.

Performance enhancements

The MongoDB CDC connector supports concurrent oplog parsing, enhancing data sync stability and reliability. The Tair (Redis OSS-compatible) connector supports asynchronous lookup joins, improving cache access efficiency and job performance.

New features

• Executing multiple DDL/DML statements in a single batch job: Create tables, perform computations, and delete tables, all in one job.

• Refreshing materialized tables on schedule: Periodically refresh historical partitions to backfill late data, ensuring eventual consistency.

• Automatically releasing idle session clusters: If new session clusters stay idle for over 30 minutes, they are automatically released to improve resource utilization.

• Blackout periods for automatic tuning: Restrict automatic resource scaling during business-critical hours, ensuring business stability. Performance tuning advice is still provided.

• Comprehensive Git integration: Support more mainstream Git tools, like Alibaba Cloud DevOps. Pull directory structure and troubleshoot with helpful error messages.

• Granular access control for data queries

Experience optimizations

• Supports console-based AI model creation, deletion, and modification. This lets you better manage AI models in the Catalogs page.

• Displays CU usage across hours for batch jobs. This metric better reflects the performance of batch jobs.

• Supports workflow fuzzy search by name in the Workflows page.

• Supports console-based Iceberg catalog creation.

API

This release includes two new APIs, two deprecated APIs, and two bug fixes. To use the new APIs, upgrade your cluster and update the pom dependency to version 1.8.0.

Previously, APIs related to Resource and DeploymentTarget could not manage hybrid billing workspaces. This release upgrades these APIs:

• Introduces new APIs:

  • CreateDeploymentTargetV2

  • UpdateDeploymentTargetV2

• Deprecates APIs:

  CreateDeploymentTarget and UpdateDeploymentTarget are deprecated. Transition to the new APIs as soon as possible.

• Resource object enhancement: Add new fields to support the configuration of hybrid billing workspaces.

• The createDeploymentDraft and modifyDeploymentDraft APIs are optimized to fix the issue that the maximum number of labels was not validated.

• The listDeployments API is optimized to validate the input of the sortName and sortOrder parameters: Only these strings are allowed: letters (a-z, A-Z) and underscores (_).