Ververica Runtime (VVR) 8.0.3, introduced in this release, occasionally causes data loss in specific scenarios, affecting data accuracy. Alibaba Cloud has announced End of Support (EOS) for VVR 8.0.3. Upgrade to VVR 8.0.5 or later as soon as possible. For upgrade instructions, see Upgrade the engine version of deployments.
This release of Realtime Compute for Apache Flink includes engine updates, platform features, connector enhancements, performance improvements, and bug fixes.
Rollout: The upgrade is deployed via a canary release strategy over a two-week period. Once the upgrade reaches your region and account, the new engine version is available for your deployments. For details, see Upgrade the engine version of deployments.
Engine updates
VVR 8.0.3 is released, built on Apache Flink 1.17.1. This version improves interoperability between Realtime Compute for Apache Flink and other Alibaba Cloud services, optimizes the MaxCompute, StarRocks, ApsaraDB for Redis, Simple Log Service, and ApsaraMQ for RocketMQ connectors for better performance and stability, and improves the usability of the CREATE TABLE AS statement.
DataStream APIs for the connectors listed above are planned for the next release.
Platform features
| Feature | What's new |
|---|---|
| Optimized automatic tuning | The stable strategy is now available in Autopilot. The system adjusts resource configuration only when a better option is identified for the entire deployment runtime—preventing unnecessary modification caused by sudden traffic bursts and helping deployments reach a stable convergence state. See Configure automatic tuning. |
| Scheduled tuning plans | Configure multiple scheduled tuning plans for a single deployment. Switch the active plan without canceling the deployment. Each plan supports multiple scheduled tuning strategies. See Configure automatic tuning. |
| Data lineage for Flink SQL deployments | View field-level data lineage for Flink SQL deployments. Find all deployments that use a specific field in a table, and trace field-level relationships between source and result tables. See View data lineage. |
| Dynamic parameter updates | Update the Parallelism parameter and specific runtime parameters of a deployment without canceling it. See Dynamically update the parameter configuration for dynamic scaling. |
| SQL optimization suggestions | View optimization suggestions and potential risks directly in the syntax check results for SQL deployments. See Develop an SQL draft. |
| Label-based search | Assign labels to SQL, JAR, or Python deployments and search by label key or label value on the Deployments page. See Create a deployment. |
| Enhanced deployment sorting and filtering | Sort deployments by health score or latency, and filter by the user who made the last modification. Available on the Deployments page. |
| Expanded logon options | Log on to the management and development consoles using a RAM role, resource directory member, or CloudSSO user—in addition to Alibaba Cloud accounts and Resource Access Management (RAM) users. See Permission management, Authorize an account to perform operations in a namespace, and Authorize access to Management Console. |
Connector and engine enhancements
| Feature | What's new |
|---|---|
| MaxCompute: complex data types and exclusive Tunnel | Read and write JSON, ARRAY, MAP, and STRUCT data types in MaxCompute. Assign an exclusive MaxCompute Tunnel to the connector for dedicated connectivity. See MaxCompute connector. |
| ApsaraDB for Redis: ALL cache policy | Set the cache parameter to ALL when creating a dimension table with the ApsaraDB for Redis connector, improving lookup performance. See ApsaraDB for Redis connector. |
| ApsaraDB RDS for MySQL: TDE support | MySQL catalogs now support ApsaraDB RDS for MySQL instances with transparent data encryption (TDE) enabled. |
| Data Lake Formation (DLF) catalogs: OSS-HDFS endpoints | Set the oss.endpoint parameter to an OSS-HDFS endpoint in a DLF catalog. See Hudi and Manage DLF catalogs. |
| StarRocks: type normalization in CREATE TABLE AS | Enable type normalization mode for schema changes when using the StarRocks connector with CREATE TABLE AS. When a source field changes to a compatible type with a different length, synchronization continues without interruption. |
| Hologres: DATE partition keys | Use a DATE column as a partition key in Hologres result tables. The partitioned table is created automatically. See Hologres connector. |
| ApsaraMQ for RocketMQ 5.x | Read from and write to ApsaraMQ for RocketMQ 5.x tables. See ApsaraMQ for RocketMQ connector. |
| Improved mini-batch and join performance | This release optimizes the mini-batch feature and join performance. In Change Data Capture (CDC) scenarios, or after deduplication and aggregation operations, throughput can increase by up to 130%. |
Fixed issues
-
MaxCompute catalog write failure on partitioned tables: Fixed an error that occurred when using a MaxCompute catalog to write data to a partitioned table.
-
Hologres dimension table join performance: Fixed poor performance when joining Hologres dimension tables.