VVR 11.1 Release Notes & Updates - Realtime Compute for Apache Flink

This release notes page covers the major feature changes and bug fixes in Realtime Compute for Apache Flink released on June 4, 2025.

Important

The upgrade rolls out gradually. Check the latest announcement in the Realtime Compute for Apache Flink console. New features in this version are available only after the upgrade completes for your account. To request an expedited upgrade, submit a ticket.

If you are upgrading, see Before you upgrade for incompatible changes that may affect your jobs.

Overview

This release includes platform updates, a new engine version, connector enhancements, and performance and security improvements.

Platform updates

Platform updates focus on developer experience and operations efficiency. Highlights:

Git integration: Bidirectional synchronization of SQL code and job configurations with GitHub, GitLab, and Gitee remote repositories, with built-in conflict resolution.
Simple Log Service (SLS) integration: Job startup logs, runtime events, and resource usage data are persisted to SLS for historical querying and auditing.
Job log configuration: Log output is now configurable through a visual interface, reducing the risk of misconfiguration.
CloudMonitor alert notifications: Alert notifications now display job names, making it faster to identify the affected job.

Engine updates

Ververica Runtime (VVR) 11.1 is now generally available. Built on Apache Flink 1.20.1, VVR 11.1 includes the following new features, enhancements, and fixes.

Before you upgrade

Before upgrading, review Upgrade the engine version of a deployment and the VVR version numbering strategy.

Incompatible changes

Java SDK — JDK 11 required: JDK 11 is now the standard runtime environment, and JDK 8 support has ended. If your JAR jobs were compiled with JDK 8, recompile and repackage them using JDK 11 before upgrading. SQL jobs are not affected.
Hologres connector — architectural changes: The Hologres connector has undergone architectural optimizations. Some connector options have been modified or removed. Before upgrading jobs that use the Hologres connector, check Connector options (VVR 11 or later) for the updated list of options.

New features

Real-time vectorization and inference powered by large language models (LLMs) The integration introduces model data definition language (DDL) statements and the ML_PREDICT function for AI inference in real-time computation. See Model DDLs for DDL syntax, ML_PREDICT for function reference, and Integrate with Alibaba Cloud Model Studio for the setup guide.
- Vector construction: Integrates with Alibaba Cloud Model Studio to vectorize streaming data in real time, supporting real-time feature computation for recommendation systems and intelligent search.
- Text inference: Generates summaries, translations, and other content using pre-trained models. Supports intelligent text processing for business use cases.

Capability enhancements

Scheduled backfill for materialized tables: Materialized tables now support creating workflows for periodic scheduling and historical data backfilling.
Real-time lookup joins with StarRocks dimension tables: Enables complex real-time analytical scenarios using StarRocks as a dimension table source.
SLS CDC and SLS-to-Paimon schema evolution: YAML Change Data Capture (CDC) now supports SLS data sources and automatic schema evolution from SLS to Paimon, simplifying metadata management in data lakehouse scenarios.
Flink CDC upgraded to version 3.4: Adds support for Flink CDC 3.4 core capabilities with enhanced data capture.
SLS data ingestion via YAML: Supports ingesting data from SLS and evolving schemas to Paimon.
MaxCompute data ingestion via YAML: Supports ingesting up to a terabyte of data into MaxCompute.
UNION ALL field alias retention: Flink SQL now retains field aliases in draft debugging results when the query contains UNION ALL.

Performance improvements

PostgreSQL CDC connector — faster snapshot reads: An optimized data segmentation mechanism with asynchronous segmentation and parallel reading significantly improves snapshot reading performance.
Tair (Redis OSS-compatible) connector — time-based cache control: The new cacheReloadTimeBlackList option lets you disable caching during specific time periods daily, helping prevent resource contention during peak hours.

Security enhancements

Tair (Redis OSS-compatible) connector — TLS/SSL support: The connector now supports TLS/SSL connections via the cacert.filepath option. This is particularly useful in security-sensitive industries such as finance and government.
Apache Parquet upgraded to 1.15.1: Fixes a remote code execution (RCE) vulnerability caused by data deserialization in the Parquet format.
MaxCompute identity reporting: The service now reports applicationName to the MaxCompute server, enabling accurate tracking of resource usage by Flink jobs and improving auditability of cross-system resource access.

Experience optimization

Input/output volume tracking: Source and sink operators now report input and output volumes, making it easier to verify data integrity and identify performance bottlenecks.

Features

Feature	Description	References
Integrate with Git repositories	Bidirectional synchronization of SQL code and job configurations with GitHub, GitLab, and Gitee, with built-in conflict resolution.	Integrate with Git (public preview)
Deliver messages to SLS	Persists job startup logs, runtime events, and resource usage data to Simple Log Service (SLS) for historical querying and auditing.	Deliver messages
Job log configuration	Configure log exports to SLS, OSS, or Kafka through the console with a visual interface.	Configure parameters to export logs of a deployment
Ingest data into MaxCompute via YAML	Use the MaxCompute connector as a sink in YAML-based data ingestion jobs, supporting up to a terabyte of data.	MaxCompute
Data ingestion optimization	The `converter-after-transform` option now supports the `FIELD_NAME_LOWER_CASE` converter type, automatically converting uppercase field names in source tables to lowercase during the cleansing phase.	Flink CDC
Elasticsearch connector optimization	Null values can now be ignored when updating Elasticsearch tables, improving robustness of data writes.	Elasticsearch
Hologres connector optimization	Architectural optimizations have modified or removed some connector options. Check the version-specific reference before upgrading Hologres connector jobs.	Hologres real-time data warehouse Hologres connector options (VVR 11+) Hologres connector options (VVR 8 or earlier)
MongoDB connector optimization	The new `ignore.delete-events.enabled` option filters MongoDB `delete` events during change data capture, reducing synchronization load.	MongoDB
MySQL connector optimization	The default value of the `property-version` option is changed to `1`.	Manage MySQL Catalog
Kafka connector optimization	The new `canal-json.infer-schema.strategy` option configures schema parsing policies for Canal JSON data — based on JSON data or `sqlType` fields. The new `json.decode.parser-table-id.fields` option generates table schema fields by parsing JSON data. The new sink-specific `debezium-json.include-schema.enabled` option checks whether Debezium JSON messages contain schema information.	Kafka
ApsaraMQ for RocketMQ connector optimization	The new `deliveryTimestampMode` and related options enable scheduled message delivery with flexible triggering rules, providing fine-grained control for time-series workloads.	ApsaraMQ for RocketMQ
Tair (Redis OSS-compatible) connector optimization	Supports reading a hashmap with multiple values per key when `cache` is set to `ALL`. The new `cacert.filepath` option enables TLS/SSL encryption for data links. The new `cacheReloadTimeBlackList` option disables caching during specified time periods daily.	Tair (Redis OSS-compatible)
StarRocks connector optimization	Supports compatible column type conversions. Supports real-time lookup joins with StarRocks dimension tables for complex analytical scenarios.	StarRocks Supported connectors
Paimon connector optimization	Supports writing and consuming `VARIANT` data through the PARSE_JSON and TRY_PARSE_JSON built-in functions, improving the performance of querying and processing JSON data.	Read and write Paimon data
AnalyticDB for PostgreSQL connector optimization	The `writeMode` option now supports `COPY` mode.	AnalyticDB for PostgreSQL
Materialized table optimization	SQL queries and connector options for materialized tables can now be modified after creation. Workflows can now be created for materialized tables to enable periodic scheduling and data backfilling.	Manage materialized tables Manage workflows
Flink SQL support for Hive Kerberos authentication	Access Hive data in a Kerberized cluster through SQL, with bidirectional identity verification and encrypted transmission.	Register a Kerberized Hive cluster Quick start with Flink SQL jobs
PyFlink Docker image upgrade	The base Docker image for PyFlink is upgraded to improve compatibility with different Python and glibc versions.
Window function optimization	SESSION WINDOW now follows Apache Flink 1.20 behavior, removing the coupling with aggregation statements that was required in VVR 8.x. Upgrade to the latest version to use the improved SESSION WINDOW.	Window aggregation
SelectDB connector — now built-in	The SelectDB connector has graduated from public preview and is now a built-in connector in Realtime Compute for Apache Flink.	SelectDB
Table API job optimization	Built-in functions can now be called in Table API queries.	Supported functions
LLM-powered real-time vectorization and inference	Introduces model DDL statements for defining AI models in Flink SQL. Adds the `ML_PREDICT` function for AI model inference in real-time computation. Integrates with Alibaba Cloud Model Studio.	Model DDLs ML_PREDICT Integrate with Alibaba Cloud Model Studio

Major bug fixes

Connectors

SLS connector — null values on escape characters: Fixed null values being returned when consuming escape characters using the new architecture.
ApsaraMQ for RocketMQ connector — message loss in delayed topics: Fixed message loss in delayed topics.
Tair (Redis OSS-compatible) connector — concurrent write conflicts: Fixed concurrent write conflicts when writing to two Tair databases simultaneously.
Hologres connector — NullPointerException on lookup joins: Fixed NullPointerException errors during lookup joins with a Hologres dimension table.
Paimon connector — IllegalStateException on primary key writes: Fixed IllegalStateException errors when writing to a Paimon table with a primary key.
Lindorm connector — lookup joins return only one row: Fixed lookup joins with a Lindorm dimension table returning only a single matching row.

SQL and transformation

YAML transformation — Calcite parser rejects string comparison operators: Fixed compilation errors caused by the Calcite parser not recognizing string comparison operators.
CTAS — schema consolidation conflicts: Fixed schema consolidation conflicts during CREATE TABLE AS SELECT operations.
Hive catalog — access denied for non-Hive tables: Fixed access being denied for non-Hive tables in a Hive catalog.

Stability and performance

Paimon sinks — slow close times: Fixed excessive close times for Paimon sink operators.
Table filtering exceptions: Fixed table filtering exceptions. As a workaround for earlier VVR versions, use the debezium.table.exclude.list option.
MiniBatch — data inconsistency: Fixed data inconsistency caused by MiniBatch processing.
PyFlink Table API — built-in function incompatibilities: Fixed incompatibilities between PyFlink Table API and Realtime Compute for Apache Flink built-in functions.