Data node release notes (2026)

Category

Description

Enhancement

Improved vector index query performance.

Bug fixes

Fixed several stability issues in AISearch.

Fixed an issue in Enterprise Edition where point-in-time recovery of instances was affected by a TSO truncation defect.

Fixed an issue in Standard Edition where descriptions did not take effect when Fully Encrypted Database rules were created or edited.

Fixed several security vulnerabilities.

Category

Description

New features

AS OF queries now support cross-DDL operations.

Vector indexes now support high-concurrency write operations.

Bug fixes

Fixed an issue where rebuilding a table with a modified primary key could cause an instance crash in edge cases.

Fixed an issue where a manual leader switch was incorrectly blocked when consensus_disable_election was enabled. Operator-initiated leader switches are no longer affected by the automatic election disable setting.

Fixed an issue where a leader switch was incorrectly reported as failed due to a brief replication lag spike on the target node.

Fixed an issue where OPTIMIZE TABLE caused deadlocks on the standby database during XA transactions with parallel replication.

Fixed a crash that could be triggered by vector indexes during ALTER TABLE operations.

Category

Description

New features

Added support for HNSW vector index.

Added support for secondary index prefix compression.

Introduced AI functions.

Added support for column encryption.

The GPP feature for accelerating table lookups on secondary indexes now supports filesort and partitioned table paths.

Bug fixes

Fixed a performance bottleneck related to XA COMMIT log application on the standby database.

Category

Description

New features

DNs now support the MODIFY_ON_UPDATE(new_value) function.

Bug fixes

Fixed an issue where proactively disabling elections during network instability could lead to a leaderless state.

Category

Description

Improvements

Added automatic shrinking for the X-Protocol query string buffer. Idle memory is freed when buffer capacity exceeds a threshold, reducing per-session memory usage for persistent connections.

Category

Description

New features

Introduced the Hippo Index format for non-unique secondary indexes. This new format significantly reduces table lookups to the clustered index during purge operations, Multi-Version Concurrency Control (MVCC) visibility checks, and implicit lock validations. This change improves query performance in write-intensive workloads and scenarios with many secondary indexes.

Refactored the consensus module to improve stability and scalability.

Introduced a pre-allocated memory pool for the FIFO Cache. This eliminates severe page fault issues in high-throughput scenarios, significantly improving memory efficiency and stability.

Index Merge and Multi-Range Read (MRR) now support the GPP feature, reducing extra page lookups and improving complex query performance.

Bug fixes

Fixed an upstream bug where an incorrect xid_t comparison rule could cause issues with transaction ordering or visibility checks.

Category

Description

New features

Added support for AliOS 8.

Bug fixes

Fixed an upstream MySQL bug where a crash might occur when using pessimistic locking to update a table after an INSTANT DROP COLUMN operation.

Fixed an upstream MySQL bug that caused the IN operator to run in O(N²) time. Complexity is now O(N*logN).