TXSQL Kernel Version | Community Version | Description |
8.4 | 8.4.6 | New features The event scheduler supports binlog replication. |
TXSQL Kernel Version | Community Version | Description |
20260430 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: New features Merged a large number of bugfix patches from the official 8.0.33 - 8.0.44 releases. An AI query optimizer has been added, which supports automatic optimization and stable continuous operation of workload-level query plans. The SQL Plan Management (SPM) feature has been added to ensure the stability of execution plans and support baseline plans, online/offline plan evolution, and automatic capture and persistence. The OPT_ESTIMATE hint for statistical information injection has been added. It supports injecting row quantity estimates for tables, indexes, and joins into the optimizer to improve the accuracy of cost estimation. The feature for Hint extraction and plan replay has been added, which can extract and fully replay historical plans. Multiple Optimizer Hints have been added: Scan method Hint: INDEX_SCAN / RANGE_SCAN / REF_SCAN. Join order Hint: Leading. Join algorithm Hint: hash_join / nested_loop_join / sort_merge_join. The show txsql_engine_version() function has been added to view the kernel engine version. SPM-related Information Schema views have been added: TXSQL_SPM_PLAN_DETAIL, TXSQL_SPM_EVOLVE_INFO, TXSQL_SPM_EVOLVE_RESULT. Performance optimization txsql_fast_path_mode is enabled by default to improve query execution performance. Optimize the memory management of InnoDB dictionary objects in scenarios with a large number of tables to alleviate memory inflation issues. Bug fixes Fixed the issue where a Replica crashes when a table without a primary key uses HASH_SCAN to replay row-based binlog due to multiple updates to a FULLTEXT index column. Fixed the OOM issue caused by excessive memory usage of dictionary objects in scenarios involving queries on a large number of tables. Synchronized 155 critical bugfixes from the official 8.0.33 - 8.0.44 releases. |
20250501 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: Bug fixes Fix the issue where a large number of concurrent before commit call failures could cause hot update transactions to get stuck. |
20250430 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: New features Add the resize bp status value. This feature adds two status values to display the progress and status of the online buffer pool resize. Add the GIS transformation function. This feature extends the MySQL spatial function ST_Transform, enabling it to transform geometries between projected coordinate systems (Projected SRS) and geographic coordinate systems (Geographic SRS). binlog is not recorded for the FLUSH TABLE operation on read-only instances. Note: binlog is not recorded for operations such as REPAIR TABLE, OPTIMIZE TABLE, ANALYZE TABLE, and FLUSH XXX (including flush table and flush status). Parallel query supports limit pushdown. When a query with a limit clause is executed in parallel, the worker sends all results to the coordinator if the limit is not pushed down. Enhance the optimizer trace to display the statistical information algorithm. Enhance the slowlog to display the statistical information algorithm. Bug fixes Fixed the issue where the PARTITION_DESCRIPTION of a secondary partition was incorrectly described. Fixed issues related to derived condition pushdown. Fixed the issue of incorrect results in anti join. Fixed the issue where I_S.PROCESSLIST reads freed memory. Fixed the issue where using order by or limit clauses in multi-table updates causes a crash. Fixed the issue of a crash caused by system clock rollback. Fixed the issue where a kernel crash occurs when queries are executed using the prepare protocol over a JDBC connection to TencentDB for MySQL. |
20241009 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: Bug fixes Fixed the issue where Instant DDL did not correctly handle column name case sensitivity. Fixed the issue where an error 1146 "Table doesn't exist" might be reported when CTE is used to query memory temporary tables. Fixed the issue of a database crash caused by the system clock rollback. Fixed the issue of a potential database crash due to loose limits of the maximum row size in Instant DDL. Fixed the issue of a kernel crash caused by executing queries using the prepare protocol after connection to TencentDB for MySQL through JDBC. |
20241005 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: New features Supported displaying the query plan with the Iterator tree structure in JSON format for the explain command. Supported encoding query plans and displaying the query plan ID in slow logs and the explain result. Bug fixes Fixed the issue of missing sql_mode in subqueries of parallel queries. Fixed the issue of crash caused by the SIGABRT signal due to the std::out_of_range exception during InnoDB statistics collection under high concurrency. |
20241001 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: Bug fixes Adapted to the scenario of disabling session-level binlog when hotspot update is enabled. sql_bin_log cannot be modified to disable session-level binlog if a session has a hotspot update transaction. It can be modified to disable session-level binlog if a session has no such transaction. If binlog is disabled for a session, subsequent hotspot update transactions in this session will not be executed. |
20240930 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: New features Support "show full binary logs" to show the last modification time. Support the parallel execution feature of multi-stage aggregation operation and execution. Parallel Copy DDL supports partition tables. Eliminate duplicate aggregate functions. Eliminate common subexpressions. Expand subquery cache to support EXISTS/IN subqueries. Correlation statistics of multiple columns. Index pushdown pruning optimization. SQL traffic throttling supports startup loading. Performance optimization Fix query cache bugs and optimize performance. Parallelize the process of checking indexes to speed up the check during backup. Bug fixes Fix the fast cleanup AHI crash issue. Fix the issue where "drop database" can cause replication interruption during the use of the recycle bin when the database contains foreign key tables. Fix the compatibility issue when the instant metadata column of the partition table is upgraded from 5.7 to 8.0. Limit the execution of XA COMMIT/ROLLBACK on RO to prevent replication interruption caused by executing related operations. Fix the issue where TRUNCATE SUBPARTITION TEMPLATE causes a crash. Fix the issue where performing the rollback operation on the partition table causes a crash. Fix the issue where multiple same columns of "subpartition by range columns" cause a crash. Fix the issue where statistical information may be cleared after COPY DDL is executed. Fix the issue where the query rewrite plugin causes a crash when it is used in multi-statement mode. |
20230704 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: Bug fixes Fixed the issue where the rebuild table operation could cause data loss under certain extreme conditions. For details, see Bug#110706. |
20230703 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: Bug fixes Fix the memory allocation method at the memory leak location to prevent memory leaks. |
20230702 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: Performance optimization Reduced the overhead caused by frequent notifying of the exchange operator in parallel queries. Bug fixes Fixed the instance crash issue that occurred when INSTANT DDL was executed to modify the column position. |
20230701 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: Bug fixes Supported restoring TXSQL physical backups by using the community edition. Replaced all keywords in the AWR feature with TXSQL_AWR. Fixed the issue where the underlying tables failed to open when competition occurred between background threads and RENAME operations on the dict cache. |
20230630 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: New features Supported the nonblocking DDL feature. Supported recording the TP/AP load statistics of the largest GTS instance in the relay log for XA commit. Supported the option for Innodb temporary tables to be shared among parallel query worker threads. Supported using partitioned tables as parallel tables in parallel queries. Supported the flashback version query feature. Supported flashback query persistence. Supported virtual indexes. Supported the range/list secondary partitioning feature. Supported the feature of automatic relay log recovery. Supported the default algorithm of DDL. INPLACE and INSTANT were optional. Supported Fast Query Cache. Supported the conversion of partitioned tables from MyISAM to InnoDB. Supported the correlated subquery cache feature. Performance optimization Optimized the BINLOG LOCK_done lock conflict. Optimize thread pool performance. Optimize the performance regression issue caused by disabling the eq_ref cache in outer joins. Enhanced the parallel query feature: Subqueries\\derived tables are executed in parallel independently: The execution plan for subqueries\\derived tables is optimized and executed in parallel independently, separate from the primary query. Nested loop join inner table parallelism: When the outer table of NLJ is small, the inner table can be selected as the parallel table for ROLL UP parallel execution. Hash join (in memory) parallel execution: Worker threads build complete hash tables respectively and perform parallel scanning on the probe side. Parallel query supports global aggregation optimization. Parallel query supports pushing down having conditions for parallel execution. Optimize the performance jitter caused by binlog purge. Supports hotspot update merge optimization. Bug fixes Fixed the issue where an assertion failure occurred during Parallel Copy DDL transaction rollback. Fixed the issue where subqueries within HashJoin conditions were not printed in EXPLAIN FORMAT=TREE output. Fixed the issue where instance operation became abnormal after an instant add operation with a redundant format is performed. Fixed the issue where global transaction IDs rolled back after a version upgrade from an older release. Fixed the issue where index merge intersect caused incorrect query results. Fixed the issue where cross-machine statistics collection could potentially block the shutdown process. Fixed the issue where a crash might occur in a master-slave environment with the historical histogram version. Fixed the issue where check index held a large number of page locks. Fixed the issue where cross-machine histograms encountered deadlocks under concurrent DDL operations. Fixed the issue where locks were not released when cross-machine histogram tasks contained an excessive number of columns. Fixed several issues related to instant ddl. Fixed the issue where update returning caused client disconnections. Fixed the null pointer dereference vulnerability discovered during vulnerability scanning. Fixed the issue where instance runtime exceptions could occur when storage-layer table structure changes during parallel execution. Fixed the issue where using WHERE column IN (list) in prepared statements leads to performance degradation. Fixed the issue where statistical information from mysqldump logical backups could be empty upon import. Fixed the issue of primary key conflicts that occurred when using Parallel Copy DDL on tables whose modifications included auto-increment columns. Fixed the issue where the build branch of a parallel query hash join could not execute in parallel when a hash join was present. Fixed the issue where the non-parallel branch of a parallel query join could not execute in parallel when a UNION was present. Fixed two memory leaks in parallel queries. Fixed the issue where partitions exit due to empty ranges in parallel queries. Fixed the issue where Outline IN-list reports errors. Fixed the issue of a crash caused by partition_id overflow during truncate partition operations. Fixed the issue where correlated subqueries referencing worker table fields in parallel queries caused incorrect query results. Fixed the issue of obtaining an incorrect offset during parallel DDL operations. Fixed the issue where parallel DDL caused instance exceptions when a unique key is added to a column containing duplicate data. Fixed the issue of assertion failures in parallel hash join debug. Fixed the issue where NDV (Number of Distinct Values) is calculated as 0 in parallel cost estimation. Fixed the issue with JSON import precision. Fixed the Bug where the FORCE INDEX ORDER BY statement skips index dive. Fixed the issue where the official subquery plan is displayed repeatedly. Fixed the issue where the count for on-disk temporary tables does not increase. Fixed the issue where proxy change user caused deadlocks. Fixed the issue where permission verification optimization in triggers calling stored procedures caused permission checks to be skipped. |
20221221 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: Bug fixes Fixed the issue where the thread_id of events persisted from the replica to the binlog changes after log_slave_update is enabled on the replica. |
20221220 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: Bug fixes Fixed the instant ddl bug. |
20221215 | 8.0.30 | Note: Starting from MySQL 8.0.29, the query results of tables in the Information Schema will use utf8mb3 instead of utf8. Versions of Connector/Net earlier than 8.0.28 do not support utf8mb3 and will report the error "Character Set 'utf8mb3' is not supported by .Net Framework" if utf8mb3 is used. If the application uses Connector/Net, upgrade Connector/Net to version 8.0.28 or later before you upgrade the TencentDB for MySQL version. For details, see: New features The user connection status monitoring feature is supported. You can view connection monitoring by using the show detail processlist command. The update wait N syntax is supported. The syntax for the nvl(), to_number(), and to_char() functions is supported. Support the cdb_kill_user_extra regular expression. Performance optimization Optimized the implementation of the binlog rotate operation to improve binlog write speed. Optimize the startup speed of TencentDB for MySQL. Optimize binlog checksum calls to reduce unnecessary CPU performance overhead. Optimize the lock hotspot of ha_innopart::external_lock to reduce lock holding time. Optimize xa::Transaction_cache to reduce lock conflicts. Reduce the execution time of ha_innopart::clear_blob_heaps. Optimize the lock hotspot of purge threads to reduce conflicts in tasks_mutex and between threads. Optimize Buffer Pool initialization, support parallel initialization, and accelerate the initialization speed. Optimize the performance of read_only and select operations under high concurrency. Optimize permission verification for prepared statements and stored procedures. Optimize change buffer access. Avoid unnecessary calls to fil_space_get to reduce hotspot issues in extreme scenarios. Optimize gtid lock conflicts during transaction commits when binlog_order_commits is disabled. Use Lock Free Hash to optimize trx_sys mutex conflicts. Optimize the snapshot overhead of the transaction system. Optimize Writeset to improve performance. Use histograms instead of index probing. Supports Parallel DDL. Bug fixes Fixed the issue of abnormal statistics such as innodb_row_lock_current_waits. Fixed the issue of excessive memory usage in GROUP_CONCAT with GROUP by. Fixed the issue where statistics are severely underestimated for long records. Fixed the issue of syntax errors when stored procedures are parsed. Fixed the issue of optimizing flush list page release concurrency in FAST DDL. |
20221002 | 8.0.22 | Bug fixes Fixed the issue of corruption in the tablespace fseg not full list caused by purging or rollback on two blob fields. Fixed the issue of a low-probability database crash occurring when THD::SECURITY_CONTEXT is accessed during the concurrent modification of SHOW PROCESSLIST and event. |
20220831 | 8.0.22 | New features Supports dynamic configuration of the MySQL version. Transparent column encryption. When creating a table, you can specify the encryption attribute for a varchar column, and the storage layer will encrypt that column. This capability is scheduled to be productized in 2023. Resolved the issue where third-party data subscription tools become abnormal when they subscribe to internal data consistency comparison SQL. Note: After a database instance undergoes migration, upgrade, or fault recovery, the system performs a data consistency check to ensure data integrity. The database comparison SQL uses the statement mode. Some third-party subscription tools may encounter exceptions when processing statement-mode SQL. After the kernel is upgraded to this version, third-party subscription tools will no longer subscribe to the internal data consistency comparison SQL. Supports the NO_WAIT | WAIT [n] feature for DDL operations. This allows a DDL operation to be rolled back immediately if it cannot obtain the MDL lock and would otherwise need to wait, or to be rolled back if it spends the specified amount of time waiting for the MDL lock. Supports the Fast Query Cache feature. This feature is designed for scenarios with more reads than writes. It is not recommended to enable it if there are more writes than reads, if data updates are very frequent, or if query result sets are very large. Supports the enhanced mts deadlock detection feature. Supports parallel query. When you enable the parallel query feature, large queries are automatically identified. The system then leverages parallel query capabilities, mobilizes multi-core computing resources, and significantly reduces the response time for large queries. Performance optimization Optimizes the snapshot overhead of the transaction system. It adopts the Copy Free Snapshot method, removes transaction delays from the global active transaction hash, and optimizes snapshotting to determine the logical timestamp of the snapshot event. In sysbench tests, the peak performance in read-write scenarios improved by 11%. Optimizes permission verification during the prepared statement process. A global variable is used to represent the permission version number. After a prepared statement completes its preparation, it records the version number. During execution, it checks whether the permission version number has changed. If there is no permission change, it skips the permission check; otherwise, it rechecks the permissions and records the version number. Optimizes the precision of time obtained in the thread pool. Optimizes the obtaining of record offsets. It caches a record's offset for each index. When conditions are met, it directly uses the cached offset, thereby eliminating the computational overhead of calling the rec_get_offsets() function. Optimizes Parallel DDL. 1. When index fields are small, it reduces the sampling memory size, thereby decreasing the sampling frequency. 2. By using K-way merging for sorting, the number of merge-sort rounds is effectively reduced, thereby decreasing the number of I/O operations. 3. When records are read, it caches fixed-length offsets, thereby avoiding generating offsets for each record every time. Optimizes the undo log information recording logic to improve insert performance. Optimizes performance when semi-sync replication is enabled. Optimize audit performance and reduce system overhead. Bug fixes Fixed the issue where Thread_memory sometimes displays abnormal values. Fixed the issue of inaccurate timestamps in the audit batch statement scenario. Fixed issues related to second-level column modification. Fixed the issue where the statement `CREATE TABLE t1 AS SELECT ST_POINTFROMGEOHASH("0123", 4326);` caused primary/secondary replication interruption. Fixed the issue of abnormal slave retries during table-level parallel execution. Fixed the issue where executing `show slave hosts` results in a "Malformed packet" error. Fixed issues related to the recycle bin. Fixed the issue where the jemalloc allocation mechanism easily triggers OOM on ARM-based machines. Fixed the issue where truncating the pfs account table caused statistics to become invalid. Fixed the abnormal situation that occurs when child tables are restored before parent tables in the presence of foreign key constraints in the recycle bin. Fixed the issue related to sql_mode logging in the logs. Fixed the issue of low-probability exceptions when CREATE DEFINER stored procedures are executed. Fixed issues related to the official Copy Free Snapshot. Fixed the thread pool performance jitter issue. Fixed the issue where the result of a hash join+union operation could be empty. Fixed memory-related issues. |
20220401 | 8.0.22 | Bug fixes Fixed the issue where an incorrect stage variable in Parallel DDL caused a stage null pointer crash during FTS index creation. Fixed the issue that could potentially cause a crash during the process of adding a full-text index. |
20220331 | 8.0.22 | Bug fixes Fixed the issue where a crash was caused by dereferencing a wild pointer in the thread pool. |
20220330 | 8.0.22 | New features writeset parallel replication is enabled by default. The system supports extended resource groups, allowing for user-level control over IO and memory usage ratios, as well as SQL timeout policies. The system supports flashback query, enabling data retrieval at any point in time within the UNDO time range. The returning syntax for delete/insert/replace/update statements is supported, allowing the data rows affected by the statement to be returned. The system supports the extension of the row-based gtid replication feature. The system supports the transaction lock optimization feature. The recycle bin feature is enhanced to support truncate table and automatically clean up tables in the recycle bin. The system supports parallel DDL, which accelerates DDL operations that require index creation through a three-phase parallel approach. The system supports the feature for quickly modifying index columns. The system supports automatic statistics collection and cross-machine statistics collection. Performance optimization Optimize gtid lock conflicts during transaction commits when binlog_order_commits is disabled. MySQL startup time is optimized by changing the creation of rsegs from single-threaded to multi-threaded during the InnoDB startup phase. Bug fixes Fixed the issue where transactions did not end after a deadlock or lock wait followed by a connection disconnection. Fixed the issue of abnormal statistics such as innodb_row_lock_current_waits. Fixed the issue where the audit plugin reported incorrect sql types when the use database statement was not executed. Fixed the issue in the asynchronous big table deletion feature where tables smaller than innodb_async_table_size were also renamed. Fixed the issue of incorrect escape character handling in the audit plugin. Fixed the issue of rollback after rapid column modification. Fixed the issue where a crash could occur during trx_sys close when an xa scenario is involved. Fixed the issue where a crash occurs when derived tables are merged. Fixed the issue of modifying binlog_format after writeset is enabled. Fixed the 1032 issue that occurs in hash scan when a row is updated in the sequence A->B->A->C. Fixed the issue where sorting indexes may become ineffective in Prepared Statement mode. Fixed the issue where the operator that consumes materialized results could be merged into the materialized operator's return value path, causing problems in understanding and displaying the execution plan. Fixed abnormal behavior in extreme cases for asynchronous big table deletion. Fixed the issue of abnormal error message prompts when the sql filter is set. Fixed errors reported when the stored procedure syntax is parsed. Fixed the issue where historical histograms cannot be applied. Fixed the compatibility issue caused by the display of the Role column in SHOW SLAVE HOSTS (show replicas). Fixed the issue where Item_in_subselect::single_value_transformer would crash when the number of columns was incorrect. Fixed the instance crash issue caused by a memory leak during cascading updates when a child table contains virtual columns and foreign key columns. |
20211202 | 8.0.22 | New features The system supports the feature for quickly modifying columns. Added support for histogram historical versioning capability. Added support for the SQL2003 TABLESAMPLE (single-table) sampling control syntax, which is used to obtain random samples from physical tables. Added a non-reserved keyword: TABLESAMPLE BERNOULLI. Added the HISTOGRAM() function, which builds a histogram for a given input field. Added support for compressed histograms. Added support for the SQL traffic throttling feature. Added support for the MySQL cluster role configuration feature. The default role is CDB_ROLE_UNKNOWN. Added a Role column to the output of the `show replicas` command to display the role. Added support for proxy. Performance optimization Optimized the update hotspot issue caused by insert on duplicate key UPDATE. Improved the application speed of hash scan by aggregating multiple identical binlog events. When the plan cache is enabled and the thread pool mode is used, the memory usage of prepare statements for point queries is significantly reduced. Bug fixes Fixed the performance instability issue that occurs after the hotspot update optimization is enabled. Fixed the issue where the parallel scan for `select count(*)` could perform a full-table scan under extreme conditions. Fixed performance issues caused by execution plan changes that resulted from statistics being read as zero under various conditions. Fixed a bug where a query remained in the query end state for an extended period. Fixed the bug where statistics are severely underestimated for long records. Fixed the bug where an error occurs when the number of aggregate functions in the SELECT clause exceeds 255 while using the Temptable engine. Fixed the issue where column names in the json_table function are case-sensitive. Fixed the bug where window functions cause correctness issues by returning early when an expression evaluates to true. Fixed the correctness issue caused by derived condition pushdown being performed even when user variables are present. Fixed the issue where SQL filter can easily cause a crash when the Rule lacks a namespace. Fixed the QPS jitter that occurs when the thread pool is enabled under high-concurrency and high-conflict conditions. Fixed the issue where master-slave bp synchronization leaks file handles under extreme conditions (such as host file system corruption). Fixed the index mapping issue. Fixed the statistics cache synchronization issue. Fixed the crash issue caused by uncleaned information after migrated update statements or stored procedures are executed. |
20210830 | 8.0.22 | New features Added support for the preload row limit feature. Added support for the plan cache point lookup optimization feature. Added support for extending the ANALYZE syntax (UPDATE HISTOGRAM c USING DATA 'json') and for the feature of directly writing histograms. Performance optimization Histograms are used instead of index probing to reduce estimation errors and I/O overhead. This capability is disabled by default. Bug fixes Fixed the issue where statistical information may be zero during online-DDL operations. Fixed the issue where generated columns on replicas are not updated. Fixed the issue where the instance hangs during binlog compression. Fixed the issue of missing GTIDs in the previous_gtids event of newly generated binlog files. Fixed the issue where a deadlock may occur when system variables are modified. Fixed the issue where the info for the replica sql thread is displayed incorrectly in show processlist. Backported bugfixes related to hash join from the official 8.0.23 release. Backported official writeset-related bugfixes. Backported bugfixes related to the query optimizer from the official 8.0.24 release. Fixed the concurrency bug related to optimizing flush list page release in FAST DDL. Optimized memory usage when the data dictionary is upgraded for instances with a massive number of tables. Fixed the crash issue that occurs when a new primary key is created after an instant add column operation. Fixed the OOM issue caused by memory growth during full-text index queries. Fixed the issue where the TIME field shows -1 in the result set returned by show processlist. Fixed the issue where histogram compatibility may cause tables to fail to open. Fixed the floating-point accumulation error in building Singleton histograms. An issue has been fixed where replication was interrupted due to longer Chinese characters in table names when using row format logs. |
20210330 | 8.0.22 | New features Supports the primary-secondary buffer pool synchronization feature: After an HA event triggers a primary-secondary switch, the secondary database typically requires a relatively long warmup period to load hot data into the buffer pool. To accelerate the warmup of the secondary server, TXSQL supports the primary-secondary buffer pool synchronization feature. Supports the Sort Merge Join feature. Supports the FAST DDL feature. Supports the feature that allows users to query the character_set_client_handshake parameter to display its current value. Performance optimization Optimized scanning the flush list for dirty page flushing: By optimizing the dirty page flushing mechanism, performance jitter during index creation is resolved, and system stability is improved. Bug fixes Fixed the deadlock issue that occurs when the offline_mode and cdb_working_mode parameters are modified. Fixed the concurrency issue in the persistence of the max_trx_id in trx_sys. |
20201230 | 8.0.22 | New features Supports dynamically setting the thread_handling mode to either thread mode or connection pool mode. Performance optimization Optimized the BINLOG LOCK_done lock conflict to improve write performance. Use Lock Free Hash to optimize trx_sys mutex conflicts, thereby improving performance. Optimized the writing of redo log data to the disk. Optimized the initialization time of the buffer pool. Optimized the cleanup of AHI when big tables are dropped. Optimize audit performance. Bug fixes Fixed the performance jitter issue caused by cleaning up innodb temporary tables. Fixed the issue of read-only performance degradation in instances with a high number of cores. Fixed the 1032 issue caused by hash scan. Fixed the concurrency security issue of the hotspot update feature. |
20200630 | 8.0.18 | New features Asynchronous big table deletion is supported: files are cleaned up asynchronously and slowly to avoid business performance jitter caused by deleting big tables. This feature requires submitting a ticket to apply for activation. The feature of automatically killing idle tasks to reduce resource conflicts is supported. This feature requires submitting a ticket to apply for activation. The transparent data encryption feature is supported. Bug fixes Fixed the issue where failover fails due to inconsistent positions between relay_log_pos & master_log_pos. Fixed the issue of data file corruption caused by asynchronous disk flushing. Fixed the issue where fsync returns EIO and repeated attempts fall into an infinite loop. Fixed the crash issue that occurs during phrase search (phrase search) in full-text indexes under multi-byte character sets. |
TXSQL Kernel Version | Community Version | Description |
20250803 | 5.7.44 | Bug fixes Fixed the issue of allowing insertion of incompatible GEOMETRY type data when the INSERT ... SELECT statement is used. |
20250802 | 5.7.44 | Note: The OOM probability of the instance can be reduced by using this kernel minor version. It is strongly recommended to upgrade to this version or later versions. New features Reduced the number of generated memory fragments to reduce the OOM probability of the instance. |
20250730 | 5.7.44 | New features Supported displaying the query plan with the Iterator tree structure in JSON format for the explain command. Supported encoding query plans and displaying the query plan ID in slow logs and the explain result. Bug fixes Fixed the issue of missing sql_mode in subqueries of parallel queries. Fixed the issue of crash caused by the SIGABRT signal due to the std::out_of_range exception during InnoDB statistics collection under high concurrency. |
20250510 | 5.7.44 | New features Added the capability of skipping transactions with GTID when the secondary server is abnormal, reducing the probability of synchronization interruption due to unexpected operations on read-only or disaster recovery instances. Bug fixes Optimized the error log generation logic in some scenarios to reduce useless logs. |
20240331 | 5.7.44 | New features Incorporated changes from official versions 5.7.36, 5.7.37, 5.7.38, 5.7.39, 5.7.40, 5.7.41, 5.7.42, 5.7.43, and 5.7.44. Added the update returning feature. Added the display of DDL progress. During DDL operations, execute the show detail processlist command to view it.Added the default table encryption feature. Performance optimization Optimized the accuracy of SQL_TYPE auditing. Index pushdown pruning optimization. Bug fixes Fixed the issue where the insert ... returning statement caused a slave crash. Fixed the issue where the slave replay thread gets stuck due to MASTER_DELAY. Fixed the issue where setting a user-defined variable via a function causes an error in session track return. Fixed the issue where hash scan consumes excessive memory. Fixed the inconsistency issue where dropping/truncating a non-existent table behaves differently when the recycle bin is enabled versus when it is disabled. Fixed the issue where pulling binlog results in garbled characters when binlog_checksum is disabled. Fixed the Parallel Copy DDL bug. |
20230601 | 5.7.36 | New features Supported flashback query persistence. Supported DROP TABLE FORCE to implement dropping InnoDB metadata. Supported Parallel Copy DDL. Supported the LIMIT clause in subqueries. Supported the conversion of partitioned tables from MyISAM to InnoDB. Bug fixes Fixed the issue of abnormal indexes in the master-slave BP synchronization feature. Fixed the issue where killing a connection during a large transaction causes an exception. Fixed the issue of obtaining incorrect user-defined variable strings in session track. Fixed the issue where creating an index fails when parallel DDL is enabled with innodb_disable_sort_file_cache. Fixed some errors in instant modify column. |
20230115 | 5.7.36 | New features Supported the nonblocking DDL feature. Support the validate password plugin. Support storing historical deadlock information. Performance optimization Asynchronous big table deletion: Temporary tables are also filtered by innodb_async_table_size. Only tables larger than innodb_async_table_size are asynchronously deleted, which improves processing efficiency. Bug fixes Fixed the issue where creating a user with grant identified by fails and causes primary/secondary replication interruption. Fixed the issue where USED_TABLES is not correctly set for GROUP_CONCAT when the DERIVED_MERGE switch is enabled. Fixed the issue where the gtid_subset function does not correctly handle null_value. Fixed the issue where the dummy index cache does not initialize system columns. Fixed the issue where instant add column on partitioned tables exceeds the maximum number of columns. Fixed the issue where pulling binlog via canal may cause OOM. Fixed the issue where the proxy reports an error when reusing connections for different users. Fixed the incorrect return values from the proxy for row count, found rows, and database settings. Fixed the issue where error messages are incomplete in the error reports during binlog sending and receiving. Fixed the issue of abnormal calculation in pagination pushdown. Fixed the issue where m_page may become invalid after Parallel DDL creates subtrees. Fixed the issue where instant modify crashes under certain character sets. |
20220716 | 5.7.36 | New features Support the InnoDB auto-increment column persistence feature. Added support for the precise memory statistics feature. Added support for the Query-level memory monitoring feature. Added support for the recycle bin feature. Added support for the parallel DDL feature. Added support for the flashback query feature. Added support for the internal xa transaction asynchronous rollback feature. Performance optimization The asynchronous big table deletion feature is optimized. The original definition of a big table was 50 GB. Now, you can control it using the innodb_async_table_size parameter, making it more flexible. Bug fixes Fixed the issue where creating an index with alter table reports error ERROR 1878 (HY000): Temporary file write failure. Fixed the issue where the PFS memory monitoring could not display the buf/buf/pool. Fixed the issue where the returning statement might cause an exception due to permission checks in certain scenarios. Fixed the issue where the parser incorrectly handles semicolons in statements, causing errors. Fixed the issue where single quotes within audit statements were not escaped. Fixed the issue of sudden memory increase on the ARM platform. Fixed the issue where modifying binlog_format after writeset is enabled causes master-slave inconsistency. Fixed the issue where a large number of threads exiting simultaneously causes high CPU usage. Fixed bugs related to drop table partition force. Fixed the issue where a stuck binlog dump causes slow instance restart. Fixed the issue where create table like temporary table does not inherit the character set in the binlog, causing source-replica synchronization failure. Fixed the issue where show detail processlist displays illegal character information. Fixed the issue where the thread pool fails to release the thread_group lock during shutdown under certain conditions. Fixed the issue where updating the parent table at the table parallel level causes abnormal instance operation. Fixed the issue where virtual columns are incorrectly calculated on the slave. Fixed the issue where gtid_subset does not set null_value to false after a row is processed. |
20211230 | 5.7.36 | New features Synchronizing the source-replica buffer pools accelerates performance recovery after HA, reducing the recovery time by approximately 90 seconds compared to the native mode. Added the backup lock feature, which provides a lightweight metadata lock to improve service availability during backups. Performance optimization Inlined the utf8/utf8mb4 my_charpos-related functions to optimize the performance of UTF_8-related functions in read_write scenarios. Upgraded the jemalloc version to 5.2.1. Optimized the acquisition of file sequence numbers during binlog rotation. Optimized the I/O performance of semi-sync replication slaves. Optimized hash scan aggregation. Optimized the startup time for crash recovery of large transactions. |
20211102 | 5.7.18 | New features Resolved the issue where third-party data subscription tools become abnormal when they subscribe to internal data consistency comparison SQL. Note: After a database instance undergoes migration, upgrade, or fault recovery, the system performs a data consistency check to ensure data integrity. The database comparison SQL uses the `statement` mode. Some third-party subscription tools may encounter exceptions when processing `statement`-mode SQL. After an upgrade to this kernel version is performed, third-party subscription tools will no longer subscribe to the internal data consistency comparison SQL. |
20211031 | 5.7.18 | New features Supports the writeset replication feature. Performance optimization Proactively advances checkpoints to improve backup success rates. Optimized hash scan index selection. The hotspot update performance optimization supports the insert on duplicate key UPDATE operation. Bug fixes Fixed the performance instability issue that occurs after the hotspot update feature is enabled. Fixed the crash issue caused by rolling back an update operation after an instant ddl operation. Fixed the issue where the create table ... select statement does not inherit compression attributes after column compression is enabled. Fixed the issue where the show variables like 'tencent_root%' statement causes an instance to crash after the skip-grant-table option is enabled. Fixed the issue where the Query Rewriter plugin crashes in read-only mode. Fixed the 1032 issue related to hash scan on partitioned tables. Fixed the issue where the first large transaction has an sbm of 0 in mts mode. Fixed the issue where the stop slave operation hangs when slave_preserve_commit_order=ON and slave_transaction_retries=0. Fixed several bugs related to XA transactions. Fixed the issue where SQL is incorrectly assembled during show create after a json field with a default value is created. Fixed the issue where a transaction cannot be rolled back after it is blocked and the connection is disconnected. Fixed the issue where statistics for the innodb persistent method may be 0 for long records. Fixed the issue where ANALYZE TABLE may cause query accumulation during the migration to version 8.0. Fixed the issue where innodb statistics are not synchronized to the Server layer in a timely manner after being changed. Fixed the issue where statistical sampling may block writes for too long, causing a crash (Bug#31889883). Fixed the issue where the innodb statistics update process may occasionally cause reads to return zero (BUG#105224). Fixed the issue where MVCC may exhibit O(N^2) complexity behavior (Bug#28825617). Fixed the crash that occurs when a connection is released and closing a temporary table triggers binlog rotation. |
20210630 | 5.7.18 | New features Added the SHOW SLAVE DETAIL [FOR CHANNEL channel] command to display the binlog timestamp that the current slave has replayed. Supports the transaction_read_only/transaction_isolation parameters. Performance optimization Optimized the application speed of hash scan. On the slave side, the application speed of hash scan is improved by aggregating multiple identical binlog events. Bug fixes Fixed the issues with duplicate primary keys, missing columns, and excessive column length in temporary tables triggered by UPDATE statements. Fixed the issue where statistical information may be zero during DDL operations. Fixed the issue of inaccurate undo log size statistics in connection status statistics. Fixed the issue where querying the metadata_locks table causes an instance to crash. Changed `of` to a non-reserved keyword. Fixed the issue where dynamically modified version numbers appear invalid in new connections. Fixed the issue of accessing wild pointers during page_cache cleaning. Fixed the issue where executing an alter table statement may trigger the "Incorrect key file for table" error. Fixed the issue of excessive memory usage by partitioned tables. Fixed the issue where the TIME field shows -1 in the result set returned by show processlist. Fixed the issue of lock waits during XA transaction replication on slave nodes. Fixed the issue of incorrect locking on partitioned tables during equal range queries. |
20210331 | 5.7.18 | New features The returning syntax for delete/insert/replace statements is supported, allowing the data rows affected by the statement to be returned. Specifically, delete statements return the before-image data, while insert/replace statements return the after-image data. Column compression is supported: Currently, there is compression for row formats and compression for data pages. However, in scenarios where a table contains some large fields and many small fields, with frequent read/write operations on the small fields and infrequent access to the large fields, these two compression methods can lead to significant unnecessary computational resource waste during read/write operations. Column compression can compress those infrequently accessed large fields, reduce the storage space for entire rows, and improve the efficiency of read/write operations. Supports the feature that allows users to query the character_set_client_handshake parameter to display its current value. Active cleanup of page cache occupied by log files is supported: This feature proactively cleans up the page cache occupied by log files using a sliding window approach via posix_fadvise, reducing operating system memory pressure and enhancing overall system stability. Performance optimization CREATE INDEX parallelization: The CREATE INDEX process requires external merge sorting, which is time-consuming. This release introduces a parallel external merge sorting algorithm, reducing the time required for CREATE INDEX by more than 50%. Optimized scanning the flush list for dirty page flushing: By optimizing the dirty page flushing mechanism, performance jitter during index creation is resolved, and system stability is improved. Bug fixes Fixed memory leak issues. Ports the json fixes from version 8.0 to enhance the stability of json usage. Fixed the 1032 issue caused by hash scan. Fixed the concurrency security issue of the hotspot update feature. Bulk ports official gcol bug fixes. Fixed the issue where the datetime type fails to compare with strings in certain scenarios. Fixed the bug where file handles were not released in the master-slave bp synchronization feature. Fixed a bug where creating a new connection while setting offline_mode could trigger a deadlock. Fixed the crash issue caused by incorrect m_end_range settings during concurrent range queries. Fixed the issue of long update times for temporary tables in groupby json fields. |
20201231 | 5.7.18 | New features Supports the use of NOWAIT and SKIP LOCKED options with SELECT FOR UPDATE/SHARE. Supports dynamically setting the thread_handling mode to either thread mode or connection pool mode. Supports the master-slave buffer pool synchronization feature. The user connection status monitoring feature includes the following monitoring items: synchronous/asynchronous IO, memory, log volume, CPU time, lock hold duration, and more. Performance optimization Optimized the transaction subsystem to improve high-concurrency performance. Optimized the startup time for crash recovery of large transactions. Optimized the writing of redo log data to the disk. Optimized the initialization time of the buffer pool. Optimized the efficiency of utf8/utf8mb4 string operations. Optimize audit performance. Removed the restriction that gtid_purged must be set to an empty value. Backup lock optimization: Introduces three new SQL statements: LOCK TABLES FOR BACKUP, LOCK BINLOG FOR BACKUP, and UNLOCK BINLOG. Unlike the FLUSH TABLES WITH READ LOCK method, which locks the entire database and makes it unavailable for service, these statements are designed to apply lightweight locks on data, enabling data access during the backup process. Both physical and logical backups can use these statements to ensure backup consistency. Optimized the dropping of big tables. Bug fixes Fixed the issue where queries against performance_schema hang. Fixed the overflow issue in the digest_add_token function. Fixed the issue where a crash occurs due to ibuf access during the official MySQL truncate table operation. Fixed the query correctness issue caused by premature const evaluation in left join statements. |
20200930 | 5.7.18 | Performance optimization Backup lock optimization: The FLUSH TABLES WITH READ LOCK method locks the entire database and makes it unavailable for service. This version provides a lightweight data backup locking method. Optimized the dropping of big tables. The fast cleanup of adaptive hash indexes, controlled by the innodb_fast_ahi_cleanup_for_drop_table parameter, can significantly reduce the time spent cleaning up adaptive hash indexes when big tables are dropped. Bug fixes Fixed the issue where a crash occurs due to ibuf access during the official MySQL truncate table operation. Fixed the issue where cold backups cannot be started after a fast column addition. Fixed the performance degradation issue caused by the frequent release of innodb memory table objects. Fixed the query correctness issue caused by premature const evaluation in left join statements. Fixed the core issue caused by a Rule class name conflict between the sql traffic throttling and query rewrite plugins. Fixed the concurrent update issue for the insert on duplicate key update statement across multiple sessions. Fixed the issue where concurrent inserts targeting auto_increment_increment cause duplicate key errors. Fixed the issue where the eviction of innodb memory objects triggers a crash. Fixed the concurrency security issue of the hotspot update feature. Fixed the issue where upgrading the jemalloc version to 5.2.1 and enabling the thread pool triggers a coredump. Fixed the issue where the lack of error handling for fwrite results in incomplete audit logs. Fixed the issue where mysqld_safe does not print logs when started as the root user. Fixed the issue where the alter table exchange partition operation causes ddl log file growth. |
20200701 | 5.7.18 | Bug fixes Fixed the INNOBASE_SHARE index mapping error. |
20200630 | 5.7.18 | New features Supports the use of NOWAIT and SKIP LOCKED options with SELECT FOR UPDATE/SHARE statements. The system supports the large transaction optimization feature, which can mitigate issues such as primary-secondary latency and backup failures caused by large transactions. Audit performance optimization: supports the asynchronous audit feature. Bug fixes Fixed the overflow issue in the digest_add_token function. Fixed the issue where inserting a blob causes an instance crash. Fixed the issue where hash scan, when encountering an update to the same row within an event but failing to locate the record, causes primary-secondary replication interruption. Fixed the issue where queries against performance_schema hang. |
20200331 | 5.7.18 | New features Added the JSON series functions from the official MySQL 5.7.22 version. Supports the Hotspot Update feature for e-commerce flash sales scenarios. Supports SQL traffic throttling. The data encryption feature supports encryption using KMS custom keys. Bug fixes Fixed the crash issue that occurs during phrase search (phrase search) in full-text indexes under multi-byte character sets. Fixed the crash issue in the CATS lock scheduling module under high concurrency conditions. |
20190830 | 5.7.18 | New features Supports the feature of skipping and continuing parsing when binlog files are corrupted. In scenarios where both the primary instance and binlog are damaged, data can be recovered to the greatest extent in the secondary database and made available for use. Supports data synchronization from non-GTID mode to GTID mode. Users can query "user thread memory usage information" by using the show full processlist command. Supports the Quick Column Addition feature for tables. It does not copy data, does not occupy disk space or disk I/O, and allows real-time changes during business peak hours. Supports auto-increment value persistence. Bug fixes Fixed the issue where replication is interrupted when a column name in a Grant statement contains a reserved word. Fixed the issue where performing a reverse scan on a partitioned table causes SQL execution to slow down. Fixed the issue where inconsistent data in the virtual index of a primary key table causes abnormal query results. Fixed the issue where InnoDB primary key range queries cause data loss. Fixed the issue where executing DDL statements on a table with a spatial index causes a system crash. Fixed the issue where primary/secondary replication is interrupted due to file length overflow in heartbeat information when the binlog is too large. Fixed the issue where deleting an event causes other events to fail to execute on schedule. Fixed the issue of incorrect aggregate query results. |
20190615 | 5.7.18 | New features The transparent data encryption feature is supported. |
20190430 | 5.7.18 | Bug fixes Fixed the null pointer reference issue that occurs when the long text feature is used in a subquery. Fixed the primary/secondary replication interruption issue caused by Hash Scan. Fixed the issue where the I/O thread of a slave node is interrupted due to a primary database binlog switch. Fixed the crash issue caused by using NAME_CONST. Fixed the "illegal mix of collation" issue caused by character set configurations. |
20190203 | 5.7.18 | New features Asynchronous big table deletion is supported: files are cleaned up asynchronously and slowly to avoid business performance jitter caused by deleting big tables. This feature requires submitting a ticket to apply for activation. The system supports the CATS lock scheduling method. When GTID is enabled, the system supports creating and deleting temporary tables and CTS syntax within transactions. This feature requires submitting a ticket to apply for activation. Implicit primary keys are supported. This feature requires submitting a ticket to apply for activation. The system supports the feature that allows non-super users to kill other users' sessions. This feature is configured through the cdb_kill_user_extra parameter, with a default value of root@%. Enterprise-grade encryption functions are supported. This feature requires submitting a ticket to apply for activation. Bug fixes Fixed the issue where replication is interrupted due to insufficient space in the binlog cache file. Fixed the issue where fsync returns EIO and repeated attempts fall into an infinite loop. Fixed the issue where replication is interrupted and cannot be recovered due to GTID gaps. |
20180918 | 5.7.18 | New features The system supports automatically killing idle transactions to reduce resource conflicts. This feature requires submitting a ticket to apply for activation. The Memory engine is automatically converted to the InnoDB engine: When the global variable cdb_convert_memory_to_innodb is ON, the table engine is converted from Memory to InnoDB during table creation/modification. The system supports the hidden index feature. The system supports Jemalloc memory management, replacing the jlibc memory management module to reduce memory usage and improve memory allocation efficiency. Performance optimization Optimized binlog switching to reduce the lock hold time during rotation, thereby improving system performance. Improved the recovery speed during Crash Recovery. Bug fixes Fixed the issue where events become invalid due to a primary-secondary switch. Fixed the Crash issue caused by REPLAY LOG RECORD. Fixed the issue of incorrect query results caused by Loose index scans. |
20180530 | 5.7.18 | New features The system supports the SQL audit feature. Table-level parallel replication is supported. This feature requires submitting a ticket to apply for activation. Performance optimization Optimized the locks on slave instances to improve their synchronization performance. Optimized the pushdown of select ... limit. Bug fixes Fixed the issue where failover fails due to inconsistent positions between relay_log_pos & master_log_pos. Fixed the Crash issue caused by the Crash on UPDATE ON DUPLICATE KEY. Fixed the "Invalid escape character in string." error caused during JSON column import. |
20171130 | 5.7.18 | New features The system supports the information_schema.metadata_locks view for querying the granted and waiting MDL statuses in the current instance. The system supports the ALTER TABLE NO_WAIT | TIMEOUT syntax, which assigns a wait timeout to DDL operations. This feature requires submitting a ticket to apply for activation. The system supports the thread pool feature. This feature requires submitting a ticket to apply for activation. Bug fixes Calculate innodb_buffer_pool_pages_data based on bytes_data to prevent overflow of this parameter. Fixed the issue where the speed limit plugin is unavailable in asynchronous mode. |
TXSQL Kernel Version | Description |
20220303 | Bug fixes Fixed the issue where an exception occurred during the release of memory allocated by mem_strdup for row_mysql_truncate_t::file_name in asynchronous big table deletion. |
20220302 | Bug fixes Fixed the memory leak issue in sql_update.cc. |
20220301 | New features The system supports dynamic configuration of the spin cycle. You can dynamically adjust the spin cycle (0~100) using the dynamic parameter innodb_spin_wait_pause_multiplier. This parameter is intended for temporary adjustments and cannot be permanently modified through the console. The system supports the feature of printing deadlock loop information. You can enable this feature using the parameter innodb_print_dead_lock_loop_info. After it is enabled, you can view the deadlock loop information by using show engine innodb status when a deadlock occurs. Bug fixes Fixed the issue where anonymous GTID transactions are generated in memory tables after a slave restart. Fixed the issue where the upgrade failed due to missing permissions for root@localhost. Fixed the issue where monitoring variable values, such as innodb_row_lock_current_waits, are abnormal. Fixed the issue of incorrect sql type mapping in the audit plugin. |
20211030 | New features The system supports optimization for large transaction replication. Performance optimization Optimize the application speed of hash scan. Bug fixes Fixed the OOM issue caused by queries on a large number of tables. Fixed the infinite loop issue caused by setting innodb_thread_concurrency to 0. Fixed the issue where statistics show as 0 for long records. Fixed the sbm jump issue. Fixed the issue of LOCK_binlog_end_pos hang. |
20210630 | New features The system supports optimization for large transaction replication. Bug fixes Fixed the correctness issue with copying when index merge is enabled. Fixed the issue where interrupting a create table ... select statement in row mode with cdb_more_gtid_feature_supported enabled causes the interruption to be replicated. Fixed the Bug where max(id) is greater than the AUTO_INCREMENT value shown in show create table. |
20201231 | Bug fixes Fixed the 1032 issue caused by hash scan. Fixed the issue where replace into in row format causes inconsistent auto-increment values between the master and slave. Fixed the memory leak issue caused by memory not being released after allocation during SQL parsing. Fixed the issue where sql mode checks were skipped during table creation with create TABLE ... as select ... . Fixed the issue where sql mode checks are skipped when an Insert statement inserts default values. Fixed the issue where sql mode checks are skipped when an update statement with bound parameters is executed. |
20200915 | New features Supports the SQL traffic throttling feature. Performance optimization Optimized the acceleration of buffer pool initialization. Bug fixes Fixed the issue where both the primary and secondary instances hang during rename table operations. Fixed a crash issue that occurs when event_scheduler is set to disable and cdb_skip_event_scheduler is changed from on to off. Fixed the issue where the maximum number of connections for tencentroot was not included in srv_max_n_threads, causing assertion failures related to sync_wait_array. Fixed a crash issue that occurs when parallel replication is enabled for primary/secondary instances, caused by structural differences in some tables within the system libraries of other cloud services' MySQL 5.6 and Tencent MySQL 5.6. Fixed the issue where INSERT ON DUPLICATE KEY UPDATE modifies the WRONG ROW. Fixed the issue where errors occur in index_mapping. Fixed the bug where mtr fails. Fixed the issue where hash scan fails to locate a record when an update to the same row occurs within an event, causing primary-secondary replication interruption. |
20190930 | New features Users can query "user thread memory usage information" by using the show full processlist command. Bug fixes Fixed the issue of gtid gaps caused by the replication filter on the secondary database. Fixed the issue where primary/secondary replication is interrupted due to file length overflow in heartbeat information when the Binlog is too large. Fixed the "illegal mix of collation" issue caused by character set configurations. Fixed the issue of primary/secondary replication interruption caused by Hash Scan. Fixed a crash issue caused by the usage of NAME_CONST. Fixed the issue where the I/O thread of a slave node is interrupted due to a primary database binlog switch. Fixed the backup incompatibility issue caused by innodb_log_checusum. |
20190530 | Bug fixes Fixed the issue of reading dirty data in RC mode. Fixed the issue where deleting a temporary table causes replay failure on the secondary server. Fixed the deadlock issue under high concurrency. |
20190203 | New features Asynchronous big table deletion: Files are cleaned up asynchronously and slowly to avoid business performance jitter caused by deleting big tables. This feature requires submitting a ticket to apply for activation. The system supports the feature that allows non-super users to kill other users' sessions. This feature is configured through the cdb_kill_user_extra parameter, with a default value of root@%. When GTID is enabled, the system supports creating and deleting temporary tables and CTS syntax within transactions. This feature requires submitting a ticket to apply for activation. Performance optimization Optimized the replay of partitioned table replication, thereby improving the replay speed of partitioned tables. Bug fixes Fixed the issue of primary/secondary inconsistency caused by insufficient temporary space. Fixed the issue where hotspot record updates hang. Fixed the issue of abnormal Seconds_Behind_Master values under parallel replication. |
20180915 | New features The MEMORY engine is automatically converted to the InnoDB engine: When the global variable cdb_convert_memory_to_innodb is ON, the table engine is converted from MEMORY to InnoDB during table creation or modification. The system automatically kills idle transactions to reduce resource conflicts. This feature requires submitting a ticket to apply for activation. Bug fixes Fixed the crash issue caused by REPLAY LOG RECORD. Fixed the issue of primary/secondary time data inconsistency caused by decimal precision problems. |
20180130 | New features The system supports the thread pool feature. This feature requires submitting a ticket to apply for activation. slave nodes support dynamically modifying replication filter conditions. Performance optimization Performance jitter caused by drop table. Bug fixes Fixed the issue of database crash caused by authentication password strings. |
20180122 | New features The system supports the SQL audit feature. Bug fixes Fixed the integer overflow issue. Fixed the issue of errors occurring during full-text index queries. Fixed the issue of slave machine crash during replication. |
20170830 | Bug fixes Fixed the issue where binlog rate limiting fails in asynchronous mode. Fixed the issue of abnormal buffer_pool status. Fixed the issue of conflict between SEQUENCE and implicit primary keys. |
20170228 | Bug fixes Fixed the character encoding bug in drop table operations. Fixed the issue where replicate-wild-do-table fails to correctly filter databases or tables containing special characters such as decimal points. Fixed the issue where the SQL thread exits prematurely after a rotate event is generated by the secondary database. |
20161130 | Performance optimization The lock_log is split to reduce the time it occupies and improve concurrent performance. The ACK thread of the primary database is decoupled to improve response time. User threads are not allowed to use the kill feature while waiting for an ACK to prevent phantom reads. Fixed the issue where an unnecessary lock_sync lock is triggered when sync_binlog != 1. |
Was this page helpful?
You can also Contact sales or Submit a Ticket for help.
Help us improve! Rate your documentation experience in 5 mins.
Feedback