PolarDB:Data node release notes for 2025

Category

Description

New feature

Added parameters to control the behavior of MySQL redo logs.

Fix

Fixed a physical backup failure (e.g., with XtraBackup) caused by a data node performing a background TRUNCATE operation on a file.

Category

Description

New features and improvements

Introduced a dynamic slot allocation mechanism for SQL throttling rules to support an unlimited number of rule entries.

Improved the internal transaction and cleanup mechanisms by decoupling transaction slot management from the Purge process, significantly increasing Purge efficiency.

Fix

Fixed an assertion issue in the transaction retention mechanism to ensure compatibility with system clock rollbacks.

Category

Description

New features & Optimizations

Optimized the transaction submission flow and internal deadlock detection logic.

Optimized the submission performance of large transactions.

Optimized Cleanout performance.

Fixes

Fixed a pcursor cleanup issue in Panda Index that occurred in specific error-handling scenarios during a transaction rollback.

Fixed an issue where the undo log space could not be truncated using TRUNCATE after an instance restart.

Fixed an issue in Three-zone deployment instances where executing the add_learner operation could cause the instance to become unavailable if a loopback address was used for communication between nodes.

Category

Description

Fixes

Fixed an issue where importing a tablespace using IMPORT TABLESPACE failed because the maximum index prefix length (max_prefix) of the source table did not match that of the destination table.

Category

Description

New features & Optimizations

Added a feature for the asynchronous deletion of large tables, which is based on the Faster drop table optimization from the community edition.

To improve data security, the RESET MASTER command is disabled by default to prevent accidental operations.

Fixes

Fixed a performance issue by optimizing the Purge system to reduce excessive I/O generation under heavy load.

Fixed an issue where transactions on Sequence tables might not roll back correctly after an instance restart.

Fixed an issue where the gtid_purged collection could be empty on a newly recovered instance.

Category

Description

Fixes

Fixed an issue where the Purge system generated a large amount of random I/O under certain scenarios and business pressures, which degraded database system performance.

Fixed an issue where executing a DDL statement during the merge phase of the Change Buffer could cause a database crash.

Category

Description

New features & Optimizations

Added support for the high-performance multi-version index Panda Index (a B-tree index). This index eliminates gap locks for unique index constraints and significantly improves SQL execution performance in various scenarios.

Added support for the AS OF syntax in parallel queries.

Added a parameter to support custom health checks for COM_PING requests.

Added a parameter to customize the latency threshold for consistent reads on secondary databases.

Added routing support for IPv6 addresses.

Enhanced the dbms_consensus.configure_followers command. This command now supports setting properties for multiple follower nodes (read-only nodes) in a single batch operation.

Optimized the range scan operation. This operation now supports pushing down the LIMIT clause for partially sorted data.

Fixes

Fixed a compatibility issue that occurred after an upgrade from MySQL 5.7 to 8.0. The issue was caused by a change in the default value of the default_collation_for_utf8mb4 parameter.

Fixed a memory leak issue that could occur when processing XA transactions.

Fixed an issue where the Log Writer thread could hold a mutex for an extended period, which prevented other threads from acquiring the lock promptly.

Fixed a bug from the community edition that caused inconsistent metadata because the data dictionary was not updated synchronously after a DISCARD TABLESPACE operation.

Category

Description

Fixes

Fixed an issue where executing PAUSE DDL could time out.

Category

Description

Fixes

Fixed an issue where executing PAUSE DDL could time out.

Category

Description

New feature

Added support for minimized runtime memory output.

Fixes

Fixed an issue that prevented a follower node from immediately re-applying changes. This occurred after the node failed to apply the changes and was subsequently corrected.

Category

Description

Optimization

Optimized the high availability (HA) switchover process to reduce the impact on online services.

Fixes

Fixed a query inconsistency issue in transaction groups that could be caused by the cleanup of some branch transactions.