2026

Category

Description

New features and optimizations

Added support for AI Function.

Added support for the AES_256 encryption algorithm.

Added performance monitoring for Binlog Dump connections on compute nodes (CNs). This feature supports three-stage timing for bottleneck analysis and heartbeat-based idle detection.

Optimized the output of EXPLAIN OPTIMIZER.

Subqueries can now be used in the JOIN ON clause.

Fixes

Fixed a startup issue where compute nodes could block new connections for up to two hours. This was caused by the synchronous fetching of OSS file metadata for cold data archiving tables.

Fixed an issue where the adaptive global secondary index (GSI) optimization was incompatible with MPP execution plans, causing query errors.

Fixed an issue where forbidPushDmlWithHint did not read the SESSION-level ENABLE_FORBID_PUSH_DML_WITH_HINT setting.

Disabled the multi-stage DDL feature by default.

Fixed an issue where a DISTINCT window function, such as COUNT(DISTINCT x) OVER(...), was incorrectly pushed down to MySQL 8.0 storage nodes (DNs), resulting in a syntax error.

Fixed an IndexOutOfBoundsException thrown when using FORCE INDEX to specify a columnstore index on a single table.

Fixed an issue where QPS statistics on compute nodes (CNs) did not include transaction control statements such as BEGIN/COMMIT/ROLLBACK. This behavior is controlled by the ENABLE_TRANSACTION_QPS_COUNT switch, which is disabled by default for backward compatibility with QPS monitoring metrics.

Fixed an issue where querying a non-existent logical table in PARTITION_HINT mode did not correctly throw the ERR_TABLE_NOT_EXIST error.

Fixed a ClassCastException that was thrown when JSON_TABLE was used with a string constant.

Category

Description

New features and optimizations

Added support for random dispatch of SHOW statements.

Added support for CTE reuse.

You can now disable writes to specified storage nodes (DNs).

Optimized the purge mechanism.

Expanded the scope of subqueries to support JOIN ON subqueries in query statements.

Fixes

Fixed an issue where SHOW VARIABLES displayed the foreign_key_checks variable as true/false instead of the standard MySQL ON/OFF format.

Fixed an issue where DECIMAL values were returned in scientific notation under the private protocol. The behavior is now consistent with MySQL.

Fixed an issue where statements that contained newline characters and resulted in an error were not automatically rolled back.

Fixed an MDL lock leak that occurred in some concurrent scenarios that involved DML and KILL statements.

Fixed an issue where the SHOW VARIABLES statement failed to correctly return the values of some variables modified within a SESSION.

Fixed an issue where executing KILL QUERY during the initialization phase of a DDL task, before a jobId was created, could cause the DDL task to hang.

Fixed a TruncatedDoubleValueOverflowException that occurred during a modulo operation on an MD5 hash string that starts with 0e.

Fixes an unsupported syntax error caused by incorrectly pushing down DISTINCT window functions, such as COUNT(DISTINCT x) OVER(...), to MySQL 8.0 storage nodes (DN).

Fixed a primary key conflict in ddl_engine_task that was caused by multiple IdGenerator instances in the DDL engine generating duplicate Snowflake IDs within the same millisecond.

Fixed incorrect handling of the [*] wildcard in JSON_EXTRACT.

Fixes an issue where OR join conditions are incorrectly converted to SemiJoin in multi-level nested subqueries.

Fixed a concurrency safety issue in AggregateGroupByDuplicatedRemoveRule to prevent occasional NullPointerException errors during high-concurrency queries.

Fixed an issue where FOR UPDATE was not correctly persisted in a LogicalView.

Category

Description

New features and optimizations

Added support for the perf_mode=boost mode, which automatically configures relevant parameters for MPP mode.

Added support for the ALTER USER LOCK/UNLOCK syntax to enable brute-force attack prevention.

Added support for sending the SHOW command randomly.

Optimized the synchronization performance of the columnstore index (CCI).

Optimized the lock-free column type change capability, provided a logic-free multi-write execution strategy, and supported more ALTER TABLE syntax.

Fixes

Fixes an issue where FOR UPDATE was not correctly persisted in LogicalView.

Category

Description

New features and optimizations

Added support for the perf_mode=boost mode, which automatically configures relevant parameters for MPP mode.

Introduced Table Partition Splitting 2.0, which provides an in-place splitting execution strategy without logical multi-write operations.

Added support for hybrid row-based and columnar storage and queries for single tables.

Added support for baseline canary releases.

Added support for columnar sub-instances.

Added support for using RETURNING to optimize performance when modifying partition keys.

Collected Number of Distinct Values (NDV) information for all columns in the columnstore index.

Optimized GroupTopN by enabling two-phase GroupTopN by default for columnar storage.

Added support for the AES_256 encryption algorithm.

Optimized the synchronization performance of the columnstore index.

Enhanced the lock-free column type change capability by providing an execution strategy that does not require logical multi-write operations and by supporting more ALTER TABLE syntaxes.

Added support for the ALTER USER LOCK/UNLOCK syntax to enable brute-force attack prevention.

Column encryption now supports external KMS keys.

Improved the compatibility of common error codes with native MySQL, such as "Table does not exist" or "Column does not exist".

Added support for pushing down subqueries in JOIN ON clauses across databases.

Expanded the applicable scope of JOIN Reorder for columnar queries.

Optimized the performance of paginated queries that use OR conditions.

Optimized cardinality estimation for single-value JOINs.

Optimized the performance of columnar subqueries by supporting decorrelation for any subquery.

TTL (Time-to-Live) now supports cold data archiving for integer types based on specific date formats.

Optimized the performance of grouped Top-N queries.

Optimized the selection logic for columnar archive tables. Queries on TTL tables without a columnstore index are now automatically routed to the columnar archive table in read-only columnar instances.

Added the enable_partition_wise parameter to centrally manage Partition-Wise mode.

Added support for the GB18030-2022 character set.

Optimized the index selection logic for queries where a global secondary index (GSI) is the only available index.

Fixes

Fixed an issue where logout log latency did not meet expectations.

Fixed an issue where the SHOW VARIABLES statement failed to correctly return the values of some variables modified within a SESSION.

Fixed an SQL execution error that was caused by incorrect pushdown of correlated subqueries.

Fixed an issue where JOIN ROWCOUNT estimation failed for multi-table JOINs that were combined with column subqueries.

Fixed a physical SQL generation error that was caused by a subquery in a Correlate operation accessing an iterative CTE.

Fixed an occasional error in the RangeScan executor that occurred during dispatch planning.

Fixed an issue where a redundant ORDER BY clause prevented the generation of a BKA JOIN.

Fixed a data error in the columnstore index that was caused by case sensitivity issues in Modify/Change Column operations.

Fixed an issue where a single-table DELETE with FORCE INDEX could cause a "database not found" error on storage nodes.

Fixed a scale-in failure that was caused by partition changes in a columnstore index.

Fixed an issue with database name replacement when pushing down cross-database CTE statements.

Fixed an issue where an archiving task for a TTL (Time-to-Live) table could fail with a null pointer exception (NPE) if the table had a columnstore index (CCI) but no archive table was configured.

Fixed an issue where the SHOW CREATE TABLE statement failed for tables named with the match keyword.

Fixed an abnormal rollback of a Modify Column operation that was caused by a multi-stage DDL locking a table in a specific availability zone due to time zone issues.

Prevented the creation of a columnstore index on instances that do not have columnar nodes.

Fixed an error that occurred when executing EXPLAIN on a multi-table pushdown DML with subpartitions.

Fixed an error in CURDATE constant folding.

Fixed a rare logical database initialization failure that could occur during compute node startup.

Fixed an issue that prevented a columnstore index from being set to INVISIBLE.

Fixed an issue where MPP queries could return incorrect results due to shuffle operations.

Fixed a potential synchronization failure in the logging engine when processing CREATE TABLE statements for a specific table group, which was caused by specifying the collation before the charset.

Fixed an MDL deadlock detection failure.

Fixed a case sensitivity issue with columnstore index names.

Fixed the return value of get_lock when the timeout is 0.

Fixed a table creation error that occurred when the boundary value definition of a range/list partitioned table contained escape characters.

Fixed a Range Scan query error within RC transactions.

Fixed a failure in the lock-free column type change feature that occurred when adding a new column before the primary key.

Fixed an error where the recycle bin feature failed to process tables that contained a global secondary index (GSI).

Fixed a CTE pushdown error for single tables in MySQL 5.7 storage nodes.

Fixed an error that was caused by using a CTE in an INSERT SELECT statement that involved only a single table.

Fixed a potential replication interruption on read-only row store instances when a DDL operation was interrupted or rolled back.

Fixed a potential optimization error in correlation queries that involved aggregate functions.

Category

Description

Fixes

Fixed an issue where the SHOW VARIABLES statement failed to correctly return the values of some variables modified within a SESSION.

Category

Description

Fixes

Fixed an upgrade error for some earlier versions of instances.

Category

Description

New features and optimizations

Improved error messages when a user does not exist.

Enhanced the lock-free column type change capability by providing an execution strategy that does not require logical multi-write operations and by supporting more ALTER TABLE syntaxes.

Added support for expanding the number of partitions by using EXPAND PARTITIONS.

TRUNCATE TABLE and TRUNCATE PARTITION now support archiving the columnstore index (CCI).

Optimized the distribution algorithm for subpartitions to resolve an issue where all subpartitions under a single primary partition were concentrated on a single storage node.

Optimized GroupTopN. Two-phase GroupTopN is now enabled by default for the columnstore index (CCI).

Introduced Table Partition Splitting 2.0, which provides an in-place splitting execution strategy without logical multi-write operations.

Added support for configuring account-based standby reads.

Column encryption now supports external KMS keys.

Fixes

Optimized the table group assignment for the columnstore index (CCI).

Fixed an unexpected failure that occurred when adding a column to a table with a clustered index.

Fixed an issue where an EXIST subquery could return duplicate results.

Fixed incorrect results for reverse-broadcast SemiJoin on unique keys.

Fixed an issue where information_schema.tables was missing the DATA_FREE statistics for the global secondary index (GSI).

Fixed errors that occurred in some subquery scenarios.

Fixed a potential interruption in the main link of the logging engine when executing an ALTER operation on a table name that contains backticks.

Optimized the xplan index selection logic to prevent slow queries caused by the incorrect selection of a bit column index.

Fixed an issue with the set data type in columnar queries.

Category

Description

New features and optimizations

Added support for hybrid row-based and columnar storage and queries for single tables.

Improved the compatibility of common error codes with native MySQL, such as "Table does not exist" or "Column does not exist".

Optimized the index selection logic in execution plans to prevent slow queries caused by the incorrect selection of a bit column index in specific scenarios.

Optimized the performance of table lookups for global secondary indexes (GSIs).

Optimized the distribution algorithm for subpartitions to resolve an issue where all subpartitions under a single primary partition were assigned to one storage node (DN) in specific scenarios.

Fixes

Fixed an issue where the SHOW CREATE TABLE statement failed for tables named with the match keyword.

Category

Description

New features and optimizations

Added support for hybrid row-based and columnar storage and queries for single tables.

Added support for directly loading incremental files from columnar nodes, which significantly reduces the overhead of cold loading for incremental data.

Added CSV file caching for columnstore indexes (CCIs).

Added the enable_partition_wise parameter to centrally manage Partition-Wise mode.

Added support for the MySQL 8.0 sql_mode: TIME_TRUNCATE_FRACTIONAL.

Optimized the performance of table lookups for global secondary indexes (GSIs).

Optimized the distribution algorithm for subpartitions to resolve an issue where all subpartitions under a single primary partition were assigned to one storage node (DN) in specific scenarios.

Optimized query performance for grouped Top-N scenarios.

Reduced transient disconnections and jitter in service traffic during maintenance on underlying storage nodes.

Optimized the idempotency check logic for migration tasks when the Locality property of a table group is modified.

Improved the compatibility of common error codes with native MySQL, such as "Table does not exist" or "Column does not exist".

Fixes

Fixed an error in UPSERT/REPLACE/INSERT IGNORE statements on storage node (DN) version 8.0 that was caused by type conversion.

Fixed a database creation failure that occurred if the name of the storage pool contained an uppercase letter.

Fixed an issue where execution plans failed to be updated correctly after DDL operations were performed on views that contained a window function.

Fixed an execution error in INSERT ON DUPLICATE KEY UPDATE statements when a TIMESTAMP NOT NULL column was updated with a zero-value time string due to the DYNAMIC IMPLICIT DEFAULT detection mechanism.

Fixed unexpected query results when sending TIMESTAMP values with fractional parts less than 0.1. This issue was caused by a parameter parsing error in the PreparedStatement interface.

Fixed incorrect metadata when multiple DNs were scaled in for logical databases that declare specific storage nodes in their Locality property.

Fixed an issue where assigning a NULL value to a TIMESTAMP column in an INSERT statement could cause an incorrect SQL statement to be pushed down to data nodes in multi-write optimization scenarios.

Fixed a potential lag in sequence values in statistics system tables.

Fixed inaccurate connection monitoring.

Fixed an issue where the connection metric for the polardbx database could display negative values in specific scenarios.

Fixed an error during INSERT operations when the expression for a generated column contains a NULL value.

Fixed unexpected failures when performing ADD COLUMN operations on tables with clustered indexes.