Minor Version | Community Version | Description |
3.1.19.001 | 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 kernel version of TDSQL-C for MySQL. For details, see: Feature updates Supports Iterator-based EXPLAIN FORMAT=JSON: After setting explain_json_format_version = 2, execute explain [analyze] format=json to display the query plan in JSON format according to the Iterator tree structure. AWR: Supports CSV format, log compression, and log size limitation. Supports query plan encoding (Plan ID): The Query Plan ID can be displayed in EXPLAIN and slow logs. AWR supports aggregated statistics by (db, digest, planid) and adds the last_exec_timestamp field to record the last execution timestamp. Statement outline supports the hint rules recommended by the AI optimizer: A new table is added to store the hint rules generated by the AI optimizer, and the following methods are used to add, delete, and update these hint rules. The hint rules recommended by the AI optimizer support automatic disabling. Supports the range/ref/index scan optimizer hint. Supports the Leading hint, which guides the optimizer to perform joins first according to the table order specified by the Leading Hint, and then Join with the remaining tables. Supports the Join type hint: This feature supports using hints to control the join algorithm. Supports using the SQL Plan Management (SPM) feature to ensure plan stability. DDL progress query: You can view the DDL algorithm, stage, and progress by running show detail processlist or querying information_schema.DETAIL_PROCESSLIST without enabling PFS. Adds Page heat management capability and optimizes the Serverless memory AS policy. Further optimizes the purge process. AWR supports recording min/max SQL: Two columns, sql_text_min and sql_text_max, are added to the logs to record the SQL text with the minimum and maximum execution latency. Performance optimization The pipeline purge model + random traversal of secondary indexes reduces blocking during the purge phase and improves the efficiency of secondary index cleanup. Issue Fixes Fixed the issue where the sql_mode copied from the coordinator thread (thd) during PQ worker startup was modified during subquery execution, causing inconsistent plan equivalence checks and triggering rollbacks (for example, in TPC-H Q22). Fixed the issue where a large amount of redo generated within a short period entered the extra_margin but did not exit, causing DDL to be blocked at log_free_check and all user threads to be blocked. |
3.1.17.002 | 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 kernel version of TDSQL-C for MySQL. For details, see: Feature updates Remove the reserved word DENSITY. Added multi-account configuration support for cdb_kill_user_extra. Added the hot restart feature for read-only instances. When a read-only instance is evicted from the cluster due to log application issues, a hot restart is performed to reset the bufferpool, various caches, and rpl_sys to catch up with the latest replication point. Added support for default values of multiple data types. The data types are BLOB, TEXT, GEOMETRY, and JSON. Default field values can now be set using literal syntax. Added support for splitting and flattening comparison expressions in ROW format into standard comparison expressions, thereby leveraging more capabilities of the optimizer. Added estimation information based on statistics to the slow logs. Added redundant range elimination. Added pushdown for limit offset calculations. Added pushdown for sum/count calculations. Added the Fast Path feature. Added the logical read-ahead feature. Added support for merging B-trees with reserved spaces to prevent frequent splitting and merging. Added abbreviated sortkey optimization. Added the concurrency control (CCL) feature. Added read/write splitting within a database proxy transaction. Added transaction resumption for the database proxy. Added the seamless upgrade feature for the CDC component. Performance optimization Optimized the issue of delayed log replay on read-only instances under increased write loads. Optimized the logic for the cdc_flink_user parameter. After the enable_default_cdc_flink_user parameter is enabled, all user behaviors are treated as flink_user. The data subscription logic of flink_user was optimized. Optimized scenarios with large IN predicates that were slow to plan and execute. In some cases, values from the IN predicate can be stored in a temporary table and optimized via a JOIN. Optimized the behavior of drop partition for global indexes. Issue Fixes Fixed the crash issue in the column compression feature when strings with a length of 255 bytes were processed. Limited the maximum configurable value of innodb_ncdb_stats_update_min_interval to 10000 to prevent crashes during startup when it was set to 18446744073709551615. Fixed the crash issue when the alter user command was executed on read-only instances. Fixed the crash issue on read-only instances caused by query_rewrite refresh. Fixed the issue where I_S.PROCESSLIST could read freed memory. Fixed the issue where a DDL operation would be permanently blocked if the creation of parallel DDL threads failed. Fixed the out-of-memory (OOM) issue during MySQL restart when a large number of partitions on a multi-partitioned table had been modified by uncommitted transactions. Fixed the issue where Instant DDL failed to properly handle case inconsistencies in column names. |
3.1.17.001 | 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 kernel version of TDSQL-C for MySQL. For details, see: Feature updates Add the reserved words DENSITY and TXSQL_CHECK. Supported parallel execution of binlog preparation and writing during the binlog flushing stage. Adaptive hash index (AHI) was supported by read-only instances. Multi-thread logical backup was supported through using MyDumper by read-only instances. Supported the capability to distribute small tables at the 1 MB level. Supported adaptive optimization of the thread pool. DDL operations using the copy algorithm on partition tables supported parallel execution. Supported the multi-column statistics feature. Supported cost-based candidate index recommendation, enabling index creation recommendations based on input SQL statements. Support the parallel execution feature of multi-stage aggregation operation and execution. Hot updates supported unique keys. Supported GB18030-2022. Audit logs supported transaction IDs for explicit transactions. Performance optimization Optimized the transaction commit process when binlog was enabled for an instance. Optimized transaction commit speed during heavy writes on full-text index tables. Optimized the refresh settings of transaction views. Optimized index dive pruning to make longer indexes easier to find. Optimizations related to column-format storage: Supported pushdown of union all/union/materialize operators. Supported pushdown of hybrid Join statements. Supported the group_concat expression. Supported creating CSI indexes on JSON columns and expanded the maximum column length supported by column-format storage. Changed the unit of the columnstore_buffer_pool_size parameter from MB to bytes. Optimized the speed of adding partitions to large partition tables. Optimized binlog commit for large transactions by removing the limitations for binlog_order_commits (binlog_order_commits = ON is no longer required strictly). Issue Fixes Fixed the issue of redo log errors in the separate buffer pool after multiple primary-secondary switches of the database. Fixed the issue where the memory allocated could be insufficient because the mlog_open_and_write_index function did not account for columns with modified order when calculating the memory required for redo. Fixed the issue that an error occurred when the innodb_backquery_enable parameter was modified on read-only instances. Fixed the issue of potential out-of-bounds memory in PFS_histogram::increment_bucket when the system clock is rolled back. Fixed the issue of an exception caused by improper handling of partition context switch during the implicit lock determination process in the scenarios involving global secondary indexes and partition tables. Fixed the issue of a potential system exception caused by global secondary index lookup not correctly switching partition context when implicit lock determination is required for the transaction rollback in scenarios involving prefix indexes and partition tables. Fixed the issue of an exception caused by performing the template-based truncate operation on a secondary partition table created without templates. Fixed the issue of a system exception caused by accidental deletion of caches of other tables with the same table name prefix during the partition table rollback, because only table name prefix matching is used for the cleanup of data dictionary caches. Fixed the issue of a potential crash that occurred when create table like was run on secondary partition tables with maxvalue. Fixed the issue of a crash caused by global indexes when the SELECT access key is the primary key, but the SELECT list does not include all partition keys. Fixed the issue of a crash that occurred when global indexes scanned records of pruned partitions. Fixed the issue of a crash in global indexes due to MurmurHash lacking character collation information. Fixed the issue that an error occurred when secondary partitions were truncated when there was only a single partition. Fixed the issue of a crash caused by global indexes after the partition tables were corrupted. Fixed the issue of a crash in the rollback path when the drop subpartition template operation on global indexes failed. Fixed the issue where the unique global index could not be created due to an inconsistency between the application order and the actual writing order of online logs. Fixed the issue of a crash caused by using uncommitted global indexes when a table is opened. Fixed the issue of a crash caused by rebuilding the primary key after an instant column was added. Fixed the issue of the memory leakage during global index rollback. Fixed the issue of an unreadable error that occurred during the rollback and was caused by existing duplicate data in the table when the DDL task of creating the unique global index was executed. Fixed the issue of a crash caused by out-of-bounds memory writes in global index tables due to copying uninitialized partition key fields from prebuilt->mysql_template. Fixed the issue of out-of-bounds reads that occurred during the copying of the subpartition template when a partition was added. Fixed the issue of garbled characters that may appear in audit logs that are queried through the frontend when users use non-UTF8 character sets but input Chinese. Fixed the issue of a crash caused by partition_id overflow during truncate partition operations. Fixed the issue of a crash caused by cleaning table caches on read-only instances when DDL operations were executed on partition tables with special characters. Fixed the issue of lock timeout caused by parallel partition adjustments. Fixed issues related to Query Cache: Query Cache supported audit. Fixed the issue where Query Cache could suffer noticeable performance degradation under low concurrency in read-write scenarios. Fixed the issue of illegal modification of thd memory, which rarely occurred, under Query Cache concurrency. Fixed the issue of a memory exception that occurred when the flush table/reset query cache operation was performed after the Query Cache memory was fully occupied. Fixed the issue of an exception caused by failed concurrent cache operations when the Query Cache cached multi-table queries. Fixed the issue where the TableName field in audit logs was displayed abnormally when recycle bin-related operations were performed. Fixed the issue where physical table names were truncated incorrectly when the results of virtual index evaluation were displayed in the JSON format. Fixed the issue of an error that occurred during the import of histograms for bigint data types. Fixed the issue of the srv_skip_dive_for_unique overflow under extreme conditions. Fixed the issue where statistical information might be cleared after the Copy DDL operation was executed. Fixed the issue of deadlock caused by the concurrent execution of change user and show processlist in the database proxy. Fixed the issue of an instance crash caused by executing show processlist. |
3.2.16.004 | 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 kernel version of TDSQL-C for MySQL. For details, see: Feature updates Added socket buffer tuning parameters for cross-region RO replication: innodb_ncdb_conn_recv_buf_size, innodb_ncdb_conn_send_buf_size, and innodb_ncdb_conn_buf_size_auto (default ON, automatically adjusted by the operating system). Added the RO-side parallel replication switch and concurrency parameters: innodb_ncdb_slave_parallel_replication, innodb_ncdb_slave_parallel_workers, innodb_ncdb_slave_scan_size, and innodb_ncdb_slave_batch_parallel_limit. This feature also supports skipping parallel distribution for excessively small log batches to avoid thread switching overhead. Added the parameter innodb_ncdb_rw_send_log_package_limit, which sets the maximum size (in MB) of log network packets sent from the RW to the RO. A value of 0 indicates no limit. Added the transaction rollback truncate undo threshold parameter innodb_rollback_trunc_threshold, which controls how many undo pages are rolled back before a truncate undo operation is triggered. Issue Fixes Fixed the issue of slow log transmission and reception caused by an undersized RO receive socket buffer. |
3.1.16.007 | 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 kernel version of TDSQL-C for MySQL. For details, see: Feature updates Added multi-account configuration support for cdb_kill_user_extra. Optimized the efficiency of undo purge. Issue Fixes Changed the default value of the parameter max_digest_length to 10240, resolving the issue where the statement outline feature was unavailable due to the truncation of digests for long SQL statements. The fix for the crash caused by single-character FTS from the community (issue#3736) has been ported. Issue: After ngram_token_size is set to 1, creating a full-text index with special characters causes a crash. Fix: The tokenization linked list is checked to ensure it does not become empty. Fixed the issue (issue#3760) where the year() function returned a longlong type. Fixed the crash issue caused by executing ALTER USER on RO. (The fix for issue#334 was incomplete.) Fixed the crash issue that occurred under extreme conditions where space was available in sync_wait_array but could not be allocated. Fixed the crash caused by a null pointer dereference that was triggered when a collation aggregation operation failed. Bug#34933045. Integrated the community fix for the issue "InnoDB: processing single character tokens with FTS parser plugin". Bug#35432973. Merged the community fix. |
3.1.16.006 | 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 kernel version of TDSQL-C for MySQL. For details, see: Issue Fixes Fixed the crash issue caused by pddl using functional indexes. Fixed the crash issue that occurred when check TABLE is executed on a gsi table. Fixed the issues related to gd18030 encoding support and encoding exceptions. Fixed the memory exception issue that may occur when a full-text index is used. Fixed the issue where the error "ERROR 1194 (HY000): Table 't1' is marked as crashed and should be repaired" was reported when the time field is queried using the COS external table feature. Fixed the issue where exporting COS external table data failed when using the mysqldump command with the parameter --txsql_cos_server_data_visiable=ON after setting set global txsql_cos_server_data_visiable=OFF. Fixed the issue where an execution error occurred when the cdc_flink_user account is used to execute the show binary logs command through a third-party client. Fixed the crash issue on read-only instances that occurred after DDL operations are performed on partition tables in a database whose name contains special characters. Fixed the issue where tables were not converted to dynamic row format after the Default row format is set to redundant when the txsql_convert_redundant_to_dynamic parameter is enabled. Fixed the issue where an infinite loop occurred after the tablespace was filled to 64T. Fixed the issue of array access out-of-bounds during system checks when there are too many connections and the reserved space of sync_wait_array is full. Fixed the issue where redo logs could be omitted during SMO operations on RTree, leading to inconsistency between the kernel and storage data. Fixed the crash issue on read-only instances that could be triggered when ACL replay is optimized. This issue occurred because a user thread, during a query, actively called handle_reload_request, which closed the table being used by the query. Fixed the issue where the select count statement was forced to use the primary key index for scanning. |
3.1.16.004 | 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 kernel version of TDSQL-C for MySQL. For details, see: Feature updates Optimize the purge process. Optimizes the undo truncate process: rollback operations for insert and temporary tables no longer write to MLOG_FENCE logs, improving parallel replication performance. Issue Fixes Fixed the crash issue that could occur when the maximum row size limit for Instant DDL is exceeded due to an insufficiently strict calculation of the limit. Fixed the crash issue that occurred during the prepare process of partition tables. Fixed the issue where transaction commits would hang and the system would print "waiting for clone gtid thread" when DML operations were performed using the Blackhole storage engine. Fixed the issue where a large amount of redo generated within a short period entered the extra_margin but did not exit, causing DDL to be blocked at log_free_check and all user threads to be blocked. Fixed the issue where a GCR read timeout would be triggered again because the lsn was not cleared after the initial timeout. |
3.1.16.003 | 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 kernel version of TDSQL-C for MySQL. For details, see: Issue Fixes Fixed issues related to the parallel query feature. Fixed the issue that a crash might occur if the optimizer mistakenly accessed a memory area that had already been reclaimed by the system when evaluating whether to use Skip Scan index scanning during parallel query execution. Fixed the issue that the Agg Pass Through component of parallel query might still take effect after a failed parallel partition adjustment, causing query result errors. Fixed the issue that a crash might occur if no parallel partitions were found in a partition during parallel query processing of partition tables. Fixed the issue that a crash might occur if the Coordinator and Worker used different transaction views in parallel query, and the Worker's transaction view expired. Fixed the issue in parallel query that parallel partitioning might hold a page lock on the current page, possibly causing a deadlock when a previous page lock was requested. Fixed the issue in parallel query that a crash might occur if the Worker incorrectly used a Query Result temporary table when the Cursor Protocol statement was executed. Fixed the issue that the parallel replication feature introduced new R-Tree logs, causing an I/O Timeout. Fixed the issue of the "report does not exist" error caused by using the TempTable engine to execute CTE queries (Bug #33814188, officially fixed). Fixed the issue of the crash caused by performing Instant/Inplace DDL on Redundant tables or the risk of being unable to pull up the instance. Optimized the issue that the read-only instance replay rate dropped severely during a large number of DCL operations. Fixed the issue that read-only instances crashed when reading abnormal data pages in some scenarios. Fixed the issue of storage burrs caused by frequent B-Tree Smo due to frequent insertion and rollback of individual Recs. Fixed issues related to Statement Outline. Fixed the issue that Statement Outline does not support Parallel Query. Fixed the issue that Statement Outline lost Warnings. Fixed several issues with EXPLAIN statement matching Outline rules. Fixed the issue of incorrect rule matching during statement truncation. Fixed the issue that the number sign comment character cannot be sent in Hints. Fixed the issue that Outline rules with a trailing semicolon cannot match. Fixed the issue of the crash of audit log and other logs caused by Cdb Outline. Fixed the issue that pending transactions were not cleaned up during Crash Recovery, causing repeated crashes. |
3.1.16.002 | 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 kernel version of TDSQL-C for MySQL. For details, see: Issue Fixes Fixed the issue of slow submissions after frequent XA transactions, where threads were stuck at waiting handler commit. Fixed the issue of read-only instance removal caused by a large number of implicit transaction rollbacks. |
3.1.16.001 | 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 kernel version of TDSQL-C for MySQL. For details, see: Issue Fixes Dynamic adjustment is supported for the Socket Buffer of the threads of the read-write/read-only instance logs. Fixed the issue where third-party subscription tools (such as Canal and Flink CDC) failed to subscribe to Binlog through read-only instances. Added parallel replication to physical logs. Fixed the issue where a large volume of write operations could cause read-only instances to be evicted. Fixed the issue that a long time was taken to modify the thread priority by calling a system function. Compatible with native data subscription behaviors based on timestamps. Fixed the issue that the log_checkpointer thread was blocked for a long time due to being unable to obtain the Log Writer Mutex. Fixed the issue that a read-only instance startup crashed due to the use of the Rewriter plugin. Fixed the issue of memory leak when the parallel query is used. Fixed the issue that column compression overflowed when the string length is 255 bytes. Merged the fixes related to Instant DDL in the MySQL community. Fixed the issue of data corruption when Instant DDL is performed with tables with foreign keys after an upgrade from MySQL 5.7 to MySQL 8.0. Fixed the issue that the backend Rollback thread crashed after Instant DDL was performed. Fixed the issue that the Redo application error occurred because Instant information is not recorded in Redo after the column position is modified in Instant. Fixed the issue that a crash occurred after an upgrade to 8029 due to the existence of Instant Add Column in the instance system table of baseline versions earlier than 8029. Fixed the issue of checking total space capacity on read-only instances. After fixing, total space capacity checks on read-only instances are canceled. Fixed the issue that write operations of other transactions were blocked by inserting large transactions into full-text indexes. Fixed the issue that the backend Optimize thread did not work when there were no read-only instances for full-text indexes. Fixed the issue of deadlock when page splitting and descending queries concurrently occurred. Fixed the issue of slow partition table opening due to logic errors in statistical information on read-only instances. Fixed the issue of deadlock caused by the concurrent execution of Change User and Show Processlist in the database proxy. |
3.1.15.006 | 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 kernel version of TDSQL-C for MySQL. For details, see: Issue Fixes Fixed issue that opening a partition table is slow due to incorrect collection of CSI statistics on a read-only instance. |
3.1.15.005 | 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 kernel version of TDSQL-C for MySQL. For details, see: Feature updates Supported two binlog subscription modes on read-only instances: 1. based on file_name and pos; 2. based on timestamp. Issue Fixes Fixed the issue that a long time was taken to modify the thread priority by calling a system function. Fixed the issue that transparent encryption failed to take effective. Fixed the issue that parallel query execution may cause page lock timeout. |
3.1.15.004 | 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 kernel version of TDSQL-C for MySQL. For details, see: Feature updates Added the query cache feature. Issue Fixes Fixed the issue where subscription tools might encounter the 'could not queue event from master' error when subscribing via cdc, under the condition that binlog checksum is enabled for some binlogs. |
3.1.15.003 | 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 kernel version of TDSQL-C for MySQL. For details, see: Feature optimization Reduced the wake-up time for Serverless instances. Issue Fixes Fixed the compatibility issue encountered by users when subscribing via flink cdc gtid, so that error scenarios are prevented during data subscription through flink. Fixed the issue where write operations of other transactions were blocked by large INSERT transactions executed during full-text index repair. Fixed the issue where the background optimize thread for full-text indexes did not work when there were no read-only instances. Fixed the crash issue that occurred during Parallel DDL execution when single-threaded sorting is used due to small index files. |
3.1.15.002 | 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 kernel version of TDSQL-C for MySQL. For details, see: Feature updates Supports TDE (Transparent Data Encryption). The secondary cache feature is supported. Issue Fixes Fixed some issues related to INSTANT DDL in the MySQL community, as detailed below. Fixed the issue of data corruption that occurs after INSTANT DDL operations are performed on tables with foreign keys following an upgrade from database version 5.7 to 8.0. Fixed the crash issue of the background rollback thread caused by executing INSTANT DDL operations. Fixed the issue where the redo application error occurred because INSTANT information was not recorded in redo after an INSTANT column position modification is performed. Fixed the crash issue that occurred after an upgrade to version 3.1.15 is performed when the instance system table from a version earlier than MySQL 8.0.29 contained an instant add column. Fixed the issue where the parallel execution plan is not displayed when the outline is opened. |
3.1.15 or 3.1.15.001 | 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 kernel version of TDSQL-C for MySQL. For details, see: Feature updates Added the recycle bin feature. Added the parallel Copy DDL feature. Added the COS external table feature (which enables databases to directly access data in COS). Added the column compression feature. Audit supports more accurate Sql_Type. Audit logs added the transaction ID and table name for the add column operation. Audit is enabled by default for newly created accounts. Supports the global index feature. Supported the range-based/list-based secondary partitioning feature. Added deadlock detection between the slave worker thread and the sql thread to prevent relay log corruption after a crash or restart. Supports the persistence of innodb update counter values. Parallel query feature updates: Parallel execution in PS mode. Hash join (in memory) is executed in parallel. derived tables from subqueries are executed separately and in parallel. The inner table of a Nested loop join is processed in parallel. Aggregation functions (std, variance, var_samp, stddev_samp) are executed in parallel. ROLL UP is executed in parallel. EXPLAIN ANALYZE is executed in parallel. Supported using partitioned tables as parallel tables in parallel queries. Supports global aggregation optimization. Supports pushing down having conditions for parallel execution. Supports rewriting statements containing subqueries into window functions when conditions are met. Supported the nonblocking DDL feature. Supports the BP isolation feature. Adds a TP/AP load monitoring feature, which allows checking the number of rows scanned by corresponding statements. Flashback queries support persistence. Subqueries in the limit clause are prohibited at the syntax level. Supports the correlated subquery result caching feature. Supports converting views into CTEs. Supports the column syntax. Supports the exchange subpartition template. Issue Fixes Fixed an issue where a binlog purge occurring during cdc binlog replenishment could cause slave binlog loss. Fixed several instant ddl bugs. Fixed an issue where invoking update returning within a stored procedure could cause client disconnection. Fixed vulnerabilities in features including sql filter and the plan equivalence judgment for parallel execution. Fixed a crash issue that could be triggered with a certain probability during parallel execution when table structures change. Fixed the performance degradation issue caused by using WHERE column IN (list) in a prepare statement. Fixed the primary key conflict issue that occurs when Parallel Copy DDL is used on table alterations involving auto-increment columns. Fixed the mutex lock conflict issue during binlog commits under high concurrency. Optimized the issue of severe performance degradation in thread pool mode when the configured thread_pool_size is small. Optimized the performance degradation issue caused by disabling eq_ref cache in outer joins. Fixed the assertion failure issue during Parallel Copy DDL transaction rollback. Fixed the issue where EXPLAIN FORMAT=TREE does not print subqueries within HashJoin conditions, an official bug. Fixed the issue where index merge intersect caused incorrect query results. Fixed the issue where cross-machine statistics collection could block the shutdown process. Fixed the issue where check index operations hold a large number of page locks. Fixed the issue where the partition process exits during parallel queries on empty ranges. Fixed the issue where the upper limit of Txsql_optimzier_context_prealloc_size overflows to 0, an official bug. Fixed the issue where the pfs_thread_setname function impacts thread pool performance. Fixed the issue of a crash caused by partition_id overflow during truncate partition operations. Fixed the issue in parallel query where correlated subqueries referencing worker table fields cause incorrect query results. Fixed the issue where an incorrect offset was obtained during parallel DDL operations. Fixed the crash issue that occurs during parallel DDL when a unique key is added to a column containing duplicate data. Fixed the issue where the debug assertion for parallel hash join fails. Fixed the issue where NDV (Number of Distinct Values) is calculated as 0 during parallel query cost estimation. Fixed the issue where index creation fails during parallel DDL when innodb_disable_sort_file_cache is enabled. Fixed the issue where the exchange id is not reset when a Prepare statement is executed in parallel query. Fixed the issue where index dive is skipped in FORCE INDEX ORDER BY statements, official Bug#34976138. Fixed the issue where the official subquery plan is displayed repeatedly, official Bug#112345. Fixed the issue where the count for on-disk temporary tables does not increase, official Bug#112556. Fixed the deadlock issue caused by the concurrent execution of change user and show processlist in the database proxy. |
3.1.14 | 8.0.22 | Feature updates Supports the columnar storage index feature (in region-level grayscale release). This feature uses columnar data format technology to store, search, and manage data, enabling better query performance and higher data compression rates. Supports enhanced parallel query features: parallel subqueries, rollup, parallel NLJ inner tables, and parallel in-memory hash joins. Issue Fixes Fixed the crash issue that may occur on read-only instances when DDL operations are performed on partition tables. Fixed the issue where garbled characters appear due to a corrupted checksum in a rotate event received during cdc subscription. Fixed the issue where precision is lost when a json string is directly inserted using '{}'. |
3.1.13 | 8.0.22 | Performance optimization Optimized the issue of slow data access caused by inefficient full table scan loops. Optimized the issue of degraded scanning performance for large queries in concurrent scenarios. Issue Fixes Fixed the CPU usage issue that occurs after the database audit feature is enabled. Fixed the issue of duplicate doc IDs in full-text indexes. Fixed the issue where full-text indexes could cause deadlocks on read-only instances in certain scenarios. Fixed the crash issue on read-only instances caused by partition table DDL operations. Fixed some official bugs. Details are as follows: Fixed the Bug related to json prepare: Bug #101284. Fixed the Bugs related to sort buffer size: Bug #32738705, Bug #33501541. Fixed the Bug related to temporary table index scan: Bug #33700735. Fixed the Bugs related to subqueries: Bug #31216115, Bug #31946448, Bug #31216115, Bug #31216115, Bug #32813547, Bug #32813554. Fixed some segmentation faults: Bug #32813554, Bug #32813547. |
3.1.12 | 8.0.22 | Feature updates You can now view storage-layer network IO traffic, time and frequency of usage, and transaction commit latency in the slow log. For details, see Query and Download Slow Logs. Database audit now supports adding the transaction ID field and table names. Supports the BLACKHOLE engine (ENGINE = BLACKHOLE;). Supports the default algorithm (inplace/instant) used for DDL execution. You can specify it using the innodb_alter_table_default_algorithm parameter. Performance optimization Optimized binlog commit for large transactions: The logs of large transactions are first written to a temporary file, which is then renamed as the next binlog, reducing the time that large transaction commits block other transactions. Optimized binlog writing: Supports converting binlogs in row format to binlogs in statement format, and supports using regular expressions to specify the databases and tables to which the optimization applies. Optimized the performance of purging binlogs, reducing performance jitter. Removed CDC, LSN, and CLUSTER as reserved keywords. Optimized the parallel DDL feature: Fixed the issue where incorrect index data is created when the number of scanning threads is greater than 1 but innodb_parallel_ddl_threads = 1. Fixed the crash issue where m_page becomes invalid after a subtree is created. Fixed the issue where error codes are not reset before index creation. Fixed the crash issue that occurs when parallel ddl creates an index with the innodb_disable_sort_file_cache parameter enabled. Fixed the crash issue that occurs when a unique key is added to a column containing duplicate data. Issue Fixes Fixed the issue where the database audit incorrectly records the ownership type of prepared statements. Fixed the issue where the database audit incorrectly categorizes the sql_type in execute statements as "other". Fixed the issue where an incorrect string is obtained for user-defined variables in session track: The return value of session track can now be constructed by using a specified string length when custom variables are obtained. Fixed the issue in parallel query where correlated subqueries referencing worker table fields cause incorrect query results. |
3.1.10 | 8.0.22 | Feature updates Read-only instances support binlog subscription. Supports the BLACKHOLE engine. Parallel query feature updates: Supports parallel queries for full table scans/full index scans/index range scans; supports variance and standard deviation functions; supports setting a parallel policy under the LIMIT syntax; and supports the Prepared Statement (PS) query mode. Supports dynamic thread pools. Supports the NOWAIT syntax. Supports flashback queries. Supports invisible indexes. Supports statement outlines. Issue Fixes Fixed the issue where instant add on partition tables resulted in too many columns. The default algorithm for ddl has been changed from instant to inplace. To use instant ddl, you must specify algorithm = instant. instant ddl is prohibited for tables that have a column named fts_doc_id. Optimized the resize buffer pool feature. Optimized the splitting logic for operator-split items in parallel queries. Fixed the performance degradation caused by the failure of the PS mode IN binary search. Optimized the query efficiency for full-table count(*) operations in parallel queries. Fixed the crash issue caused by an incorrect stage variable in Parallel DDL, which resulted in a stage null pointer crash during FTS index creation. Fixed the crash issue that could occur during the creation of a new full-text index. Fixed the issue where HA could cause gtid loss in some cases. Fixed the crash issue that could be triggered probabilistically by the show create table command after the mysqld process is killed during a ddl operation and restarted. Fixed the issue where premature release of the fts cache lock by a full-text index caused read-write inconsistency. Fixed the issue where a transaction rollback during ddl commit caused master-slave data inconsistency. Fixed the deadlock issue that occurred when a transaction is killed and a database is dropped during the process of creating a temporary table and inserting data. Fixed the issue where operations on a database with the same name as a mysql system database failed. Fixed the issue where canal could not obtain the gtid starting point using show master status when the binlog is extracted through RO. |
3.1.9 | 8.0.22 | Feature updates A new cdc capability has been added, which allows direct and repeated backtracking/extraction of binlogs from user-defined log retention periods. This resolves the issue of compute nodes losing local binlogs in scenarios such as HA. To set the binlog retention period, see configure the backup retention period. Optimize the pages purge rate to improve database performance. Issue Fixes Fixed the issue in parallel query where the worker's inability to pass the table-level NULL ROW FLAG flag to the coordinator caused incorrect results. Fixed the issue in parallel query where pushing down the sort order to the table caused the sort operator to fail to obtain the order list during operator splitting, resulting in a core dump. |
3.1.8 | 8.0.22 | Feature updates The parallel query feature has been added. It automatically identifies complex queries, leverages parallel query capabilities to mobilize multi-core computing resources, and significantly reduces the response time for large queries. For usage, see the enable or disable parallel query feature description. Issue Fixes Fixed several database issues that occurred in debug mode. Fixed the issue where abnormal information appeared in database proxy-related fields when the show detailed processlist command is used to display connection information. |
3.1.7 | 8.0.22 | Feature updates The password dictionary parameter has been added. For usage, see the custom password strength feature description. The issue of HA's dependency on dictionary files has been optimized. The kernel supports the completion of purged binlogs. The Serverless architecture supports a built-in database proxy, which enables connection persistence and the momentary disconnection prevention feature, thereby resolving the issue of errors occurring during the first connection wake-up. |
3.1.6 | 8.0.22 | Issue Fixes Fixed the crash issue caused by encountering a full-text index (non-tree index) during check index. Fixed the issue where the database proxy's liveness probe caused the kernel to output a large number of errlogs, and suppressed redundant log printing. Fixed the issue where an uninitialized gcr lsn in the session lsn tracker could cause a database proxy random value to be returned, leading to statement execution timeout. |
3.1.5 | 8.0.22 | Feature updates Added traffic throttling for bulk insert. Supports configuration via the change buffer and merge methods. Database proxy is supported. For usage, see the Database Proxy feature description. Read-only instances are supported to perform logical backups. Supports parallel replication of binlog at the table level. Added the SQL traffic throttling feature. Added Hot Update Protection. Supports interesting order determination for sort merge join. Added the TABLESAMPLE feature. Added the HISTOGRAM() function. Added the histogram version history feature and compressed histograms. Added the show detail processlist command. Performance optimization Optimized the physical replication of transactions, significantly improving write-only performance. Optimized the parallel initialization of the transaction system, reducing system startup time. Optimized the page locking logic during log replay on read-only instances to accelerate the replay thread. Issue Fixes Fixed an issue where the mysql client abnormally exits when it receives incomplete packets. Fixed the crash issue in the fsp management segment caused by nested mtr commit order errors within blobs. Fixed a transaction consistency issue that may occur when read-only instances access secondary indexes during host purge operations. Fixed the issue where the backup lock cannot be acquired due to interference from the lock table. Fixed the issue where several keywords introduced by cynosdb cannot be used as identifiers, such as CDB_GET_TABLE_VERSION, CLUSTER, and THREADPOOL. Fixed the issue where instance startup time is prolonged due to the main thread being unable to acquire the dict op lock during the rollback of large or long-row transactions at instance startup. Fixed the crash issue caused by reusing a trx after an undo page allocation failure. Fixed the crash issue caused by migrating a partition table from an extended tablespace to the system tablespace using the alter TABLE command. Fixed the crash issue that occurs during startup after an incomplete write of the truncate log. Fixed the crash issue caused by inserting data after drop table partition force is executed. Fixed the crash issue that may occur when data is updated by instant DDL and then rolled back. Fixed the issue where creating a temporary table using `create TEMPORARY TABLE like` referencing a table in an extended tablespace fails. Fixed the OOM issue caused by continuously increasing cache during data loading for full-text index tables. Fixed the issue of unstable performance after the hot update optimization is enabled. Fixed the issue where `select count(*)` parallel scanning could result in a full table scan under extreme conditions. Fixed the issue where statistics information reads zero under various conditions, and fixed the official Bug#31889883. Fixed the issue where a query remains in the query end state for an extended period. Fixed the case-sensitive issue with column names in the json_table function, official Bug#32591074. Fixed the issue where an error is reported when the number of aggregate functions in the selected columns exceeds 255 while using the Temptable engine. Fixed the bug where window functions cause correctness issues due to expressions returning early when `return true`. Fixed the correctness issue caused by derived condition pushdown being applied when user variables are present. Fixed the crash issue that occurs when the SQL filter is applied to a Rule without a namespace. Fixed the QPS jitter issue that occurs when the thread pool is enabled under high-concurrency and high-conflict conditions. Fixed the crash issue caused by information not being cleaned up after update statements or stored procedures are executed during migration. Fixed the issue in the current version where the histogram cannot be stopped using CTRL+C. |
3.1.3 | 8.0.22 | Feature updates Supports adding the binlog with the specified filename to the index file. Added the backup lock, with the syntax: LOCK TABLES FOR BACKUP, UNLOCK TABLES. Added the binlog lock, with the syntax: LOCK BINLOG FOR BACKUP, UNLOCK BINLOG. Issue Fixes Fixed the bug where dynamic metadata persistence caused instance table corruption or visibility errors. Incorporated the official bugfix Bug #32897503 to resolve the issue where some query statements under prepare statements followed an incorrect execution path. Incorporated the official bugfix: the crash issue that occurs when a resource group setting fails. Fixed the bug where the previous gtid is left blank after an HA switchover. Fixed the bug where an auto-increment column could be set to a value less than the maximum value already inserted. Fixed the issue where explicit transactions on a read-only instance block the replay thread from replaying DDL logs. Fixed the crash issue that may occur when replicating on a read-only instance after DDL operations are performed on tables whose names contain a hyphen ("-"). Fixed the crash issue that occurs when a read-only instance starts up, encounters DDL recovery, and then crashes while operating on the DDL log system table to apply for undo. |
3.1.2 | 8.0.22 | Feature updates Supports MySQL 8.0, adds read-only nodes, and enables primary/secondary physical replication. Supports expandable table spaces, with instance capacity supporting up to 1PB+. Added the preload row limit feature, which delivered a performance improvement of 1% to 5% in point query-related tests. Supports the extended ANALYZE syntax (UPDATE HISTOGRAM c USING DATA 'json') and the feature for directly writing histograms. Performance optimization Replaces index probing with histograms to reduce estimation errors and I/O overhead. This capability is disabled by default. Issue Fixes Fixed the issue where updates related to large object pages were not logged when a full-text index that contains large object columns is created. Fixed the inconsistency in the undo page format between the compute and storage layers, where the definition of TRX_UNDO_HISTORY_NODE differs. Fixed the issue where statistics information might be zero during online-DDL operations. Fixed the issue where generated columns on the replica are not updated. Fixed the issue where the instance hangs during binlog compression. Fixed the issue where the gtid is missing in the previous_gtids event of newly generated binlog files. Fixed the issue where a deadlock might occur when system variables are modified. Fixed the issue where the info field for the replica sql thread is displayed incorrectly in show processlist. Ports the hash join-related bugfixes from the official 8.0.23 release. Ports the writeset-related bugfixes from the official release. Ports the optimizer-related bugfixes from the official 8.0.24 release. Fixed a concurrency bug related to optimizing the flush list page release in FAST DDL. Optimizes memory usage when the data dictionary is upgraded for instances with a massive number of tables. Fixed the issue of a crash occurring 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 could prevent tables from being opened. Fixed the floating-point accumulation error in Singleton histogram construction. Fixed the issue where replication is interrupted when a table name contains long Chinese characters in row-format logs. |
3.1.1 | 8.0.22 | Feature updates Incorporated the official changes from 8.0.19, 8.0.20, 8.0.21, and 8.0.22. Supports dynamic configuration of the thread_handling mode, either as thread mode or connection pool mode. TXSQL supports the primary-secondary buffer pool synchronization feature: After an HA event triggers a primary-secondary switch, the secondary database typically requires a significant amount of time to warm up by loading hot data into the buffer pool. To accelerate the secondary's warm-up process, TXSQL provides this feature. Supports the Sort Merge Join feature. Asynchronous commit is supported. You can enable it by setting the parameter innodb_log_sync_method = async when the thread pool is enabled and binlog is disabled. Supports the FAST DDL feature. Supports the feature that allows users to query and display the current value of the character_set_client_handshake parameter. Supports database audit. Performance optimization Optimized the flush list scanning for dirty page flushing: By optimizing the dirty page flushing mechanism, performance jitter during index creation is resolved, and system stability is improved. Optimized the BINLOG LOCK_done lock conflict to improve write performance. Use Lock Free Hash to optimize the trx_sys mutex conflict and improve 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 large tables are dropped. Issue Fixes Fixed the deadlock issue that could occur when the offline_mode and cdb_working_mode parameters are modified. Fixed the concurrency issue with the persistence of max_trx_id in trx_sys. Fixed the performance jitter issue caused when innodb temporary tables are cleaned up. Fixed the issue of read-only performance degradation on instances with a high number of cores. Fixed the issue where hash scan causes error 1032. Fixed the concurrency security issue of the hot update feature. |
3.0.1 | 8.0.22 | Feature updates The cynos_version is supported and can be queried in three ways: select CYNOS_VERSION(), select @@cynos_version, and show variables like 'cynos_version'. A space limit parameter has been added. When the total space exceeds the limit, an expansion update will fail with an error. A warning is displayed to prompt the user to free up space or upgrade the specification. The read-only parameter innodb_ncdb_log_priority has been added, which indicates the priority of the background log thread for the primary instance. The read-only parameter innodb_ncdb_apply_priority has been added, which indicates the priority of the log replay thread for read-only instances. The dynamic parameter innodb_ncdb_fast_shutdown has been added. It controls whether a process shuts down quickly. When this parameter is enabled, the process exit will skip certain global structure deconstruction operations, reducing shutdown time. This parameter is disabled by default. The parameter innodb_max_temp_data_file_size has been added. It is a read-only parameter with a default value of 128M. When this parameter is greater than 0, it represents the maximum temporary table space that can be accommodated in local storage. |
Minor Version | Community Version | Description |
2.1.14.001 | 5.7 | Feature updates Added parallel replication to physical logs. Fixed the issue where a large volume of write operations could cause read-only instances to be evicted. Issue Fixes Fixed the issue where creating an index could cause an OOM error. Fixed the issue where the transaction ID (trx id) of implicit transactions in audit logs was displayed incorrectly. Fixed the crash issue on read-only instances caused by the query rewrite plugin. Fixed the issue of character encoding translation errors for non-UTF8 data in audit logs. Fixed the issue where minute-level DML blocking occurred during online DDL operations. Fixed the issue where MySQL bug #101827 caused an infinite loop. Fixed the issue of slow submissions after frequent XA transactions, where threads were stuck at waiting handler commit. Fixed the issue of global consistent read timeout. Fixed the issue where gtid persistence could fail when the storage space was full. Fixed the deadlock issue when cdb_store_meta_to_table was enabled and storage capacity was full, preventing DDL log writes. Fixed the issue where accessing an invalid page could cause instance crashes. Fixed the issue of slow instance shutdown after log I/O timeout. Fixed the crash issue during exchange partition operations due to uncleaned table caches on read-only instances. |
2.1.13.002 | 5.7 | Feature updates The secondary cache feature is supported. Issue Fixes Supported higher floating-point number precision for parsing JSON strings. Fixed the issue of statistical errors of innodb_ncdb_avg_log_rsp_time. Fixed the issue that the result obtained by executing the checksum table command was unstable when the table contains JSON fields.Optimized the binlog subscription feature. |
2.1.13.001 | 5.7 | Feature updates Supports the flashback query feature. Supports the returning feature. Supports the Nonblocking DDL feature. Supports the parallel Copy DDL feature. Added DDL process information. Note: During DDL execution, you can view the DDL process information by running the show detail processlist; or select * from information_schema.DETAIL_PROCESSLIST; command.Supports using LIMIT & IN/ALL/ANY/SOME in subqueries. Audit is enabled by default for newly created accounts. Audit logs added the ClientPort field. Issue Fixes Fixed the "reached open_files_limit" error caused by excessive cdc retries. Fixed the issue where single quotes within audit statements were not escaped. Fixed the JSON display precision issue. Fixed the threadpool status calculation issue. Fixed the issue where the aggregate function sum() returned incorrect results when cdb_enable_sumagg_pushdown was enabled. Fixed the issue where incorrect query results were caused by the null_value of gtid_subset not being reset. Fixed the issue where GROUP_CONCAT did not correctly set USED_TABLES when the DERIVED_MERGE switch was enabled. Fixed the issue where the database proxy encountered errors when reusing connections with different users. Fixed the error returned by the database proxy for row count, found row, and db settings. Fixed the issue of excessive memory usage during hash scan. Fixed the issue where an incorrect string is obtained for user-defined variables in session track. Fixed the issue where queries using column names with * returned errors. Fixed the null pointer issue caused by the SQL filter. Disable the returning statement when multiple tables are involved. Fixed the crash issue that occurs when the 'show processlist' command is executed. Fixed the crash issue on cross-AZ standby instances caused by the Returning syntax. Audit bug fix: Truncating multi-byte characters in long logs results in garbled log content that cannot be parsed. Audit bug fix: Audit logs frequently allocate memory when processing escape characters, resulting in poor performance. Fixed the issue where errors are reported during flink - cdc timestamp subscription. Fixed the crash issue that occurs in some environments when the tablespace name is validated and left blank. |
2.1.12 | 5.7 | Feature updates Read-only instances support binlog subscription via position. Performance optimization Optimized the performance of purge binlog, reducing performance jitter. Issue Fixes Fixed the incompatibility issue that occurs in some scenarios when Flink subscribes to the binlog of TDSQL-C read-only instances. Fixed the official Bug #27422376. Fixed the issue where read retries occur when expired data pages are encountered during B-Tree reverse traversal. Fixed the issue where pages might be repeatedly reclaimed when the buffer pool is resized down. |
2.1.11 | 5.7 | Feature updates Database audit now supports adding the transaction ID field and table names. Multi-threaded logical backup is supported. Database proxy supports transaction split. Performance optimization Optimized the memory usage mechanism to reduce the risk of OOM. Removed CDC, LSN, and CLUSTER as reserved keywords. Optimized the storage apply log logic to enhance the storage layer's capability of processing redo logs. Issue Fixes Fixed the issue where the database audit incorrectly records the ownership type of prepared statements. Fixed the issue where the database audit incorrectly categorizes the sql_type in execute statements as "other". Fixed the official Bug #111686 and Bug #26225783. |
2.1.10 | 5.7 | Feature updates Read-only instances support binlog subscription. Supports the BLACKHOLE engine. Supports resizing the buffer pool. Issue Fixes Fixed the issue where RO binlog extraction may frequently disconnect. |
2.0.23/2.1.9 | 5.7 | Feature updates Supports dynamic thread pools. Supports the NOWAIT syntax. Supports returning. Supports auto-increment column persistence. Supports invisible indexes. Supports computation pushdown. Supports buffer pool initialization. Added Hot Update Protection. Issue Fixes Fixed the issue where HA could cause gtid loss in some cases. Fixed the crash issue that could be triggered probabilistically by the show create table command after the mysqld process is killed during a ddl operation and restarted. Fixed the issue where premature release of the fts cache lock by a full-text index caused read-write inconsistency. |
2.0.22/2.1.8 | 5.7 | Feature updates A new cdc capability has been added. It allows direct and repeated backtracking/extraction of binlogs from user-defined log retention periods, resolving the issue where compute nodes lose local binlogs in scenarios such as HA. To set the binlog retention period, see configure the backup retention period. Optimize the pages purge rate to improve database performance. Issue Fixes Fixed the issue where compute instances repeatedly crash when the storage space limit is triggered. |
2.0.21/2.1.7 | 5.7 | Feature updates The password dictionary parameter has been added. For usage, see the custom password strength feature description. The issue of HA's dependency on dictionary files has been optimized. The kernel supports the completion of purged binlogs. The Serverless architecture supports a built-in database proxy, which enables connection persistence and the momentary disconnection prevention feature, thereby resolving the issue of errors occurring during the first connection wake-up. |
2.0.20/2.1.6 | 5.7 | Issue Fixes Fixed the issue where the database proxy's liveness probe caused the kernel to output a large number of errlogs, and suppressed redundant log printing. Fixed the issue where an uninitialized gcr lsn in the session lsn tracker could cause a database proxy random value to be returned, leading to statement execution timeout. Fixed the issue where creating temporary tables on read-only instances could cause deadlocks. |
2.0.19 | 5.7 | Feature updates Read-only instances are supported to perform logical backups. Database proxy is supported. For usage, see the Database Proxy feature description. Supports parallel replication of binlog at the table level. Supports configuration via the change buffer and merge methods. Added the show detail processlist command. Issue Fixes Fixed the official json character set related Bug #22991924. Incorporated the official bugfix Bug #25865525 to resolve the issue where the LOAD DATA INFILE statement failed to read escape characters followed by UTF-8 characters. Incorporated the official bugfix Bug #31529221 to resolve the issue where ALTER TABLE operations fail and report an "Incorrect key file" error. Incorporated several official bugfixes related to generated columns and cascade deletion, including Bug #33053297, Bug #32124113, and Bug #29127203. Fixed the official Bug #31599938, which addresses the issue where the system hangs when reset master is executed on a slave with log_bin disabled while binlogs are being chased. Fixed the issue where instance startup time is prolonged due to the main thread being unable to acquire the dict op lock during the rollback of large or long-row transactions at instance startup. Fixed the crash issue caused by reusing a trx after an undo page allocation failure. Fixed the crash issue caused by migrating a partition table from an extended tablespace to the system tablespace using the alter TABLE command. Fixed the crash issue that occurs during startup after an incomplete write of the truncate log. Fixed the crash issue caused by inserting data after drop table partition force is executed. Fixed the crash issue that may occur when data is updated by instant DDL and then rolled back. Fixed the issue where creating a temporary table using `create TEMPORARY TABLE like` referencing a table in an extended tablespace fails. Fixed the OOM issue caused by continuously increasing cache during data loading for full-text index tables. |
2.0.17 | 5.7 | Feature updates Supports adding the binlog with the specified filename to the index file. Issue Fixes Incorporated the official bugfix BUG#25865525 to resolve the issue where the LOAD DATA INFILE statement fails to read escape characters followed by UTF-8 characters. |
2.0.16 | 5.7 | Performance optimization Optimized undo space truncate, significantly improving the speed of undo truncate for large-scale instances. Optimized the performance of large-scale range queries on read-only instances. Issue Fixes Fixed the issue where the backup lock cannot be acquired for backup due to interference from the lock table statement. Fixed the issue where RO experiences table share corruption after an instant add operation involving thousands of columns. Fixed the issue where binlog replay fails with an error when the binlog content contains escaped keywords. Fixed the issue where an external prepared XA transaction that is not explicitly rolled back blocks the normal shutdown process. Fixed the issue where a warning "tablespace -1 not found" is reported during the startup process of a read-only instance. |
2.0.15 | 5.7 | Feature updates Supports expandable table spaces. When a single table space exceeds the innodb_ncdb_extend_space_threshold configuration, new tables are created in the extended table space. Added the following JSON functions: JSON_MERGE_PRESERVE, JSON_MERGE_PATCH, JSON_PRETTY, JSON_STORAGE_SIZE, JSON_ARRAYAGG, and JSON_OBJECTAGG. Optimized the table lock recovery process during system startup, reducing the startup time. Issue Fixes Fixed official bugfixes, which include performance issues with group by on text columns and issues related to multiple virtual columns. Fixed the crash issue that may occur when a large transaction rollback and a shutdown operation are concurrent after instance startup. Fixed the issue where an instance exits due to repeated I/O retry failures on related pages during undo space truncate retry after a failure. Fixed the crash issue that occurs when a read-only instance is shut down and statistics information update accesses the snapshot cache. Fixed the issue where the truncate log may lag behind the truncate operation during undo space truncate. Fixed the issue in the official bugfix where an error in the ICP check during partition table scanning caused slow queries. Fixed the crash issue that occurs when a forward scan in a read-only instance encounters partial replay of index split logs. |
2.0.14 | 5.7 | Issue Fixes Fixed the issue where occupied space is not reclaimed when temporary tables are deleted in a read-only instance. Fixed the issue where a read-only instance exits because it reads an old page version due to storage route changes. Fixed the crash issue that may occur when information_schema.metadata_locks is queried. Fixed the concurrency issue between forward scanning and B-tree SMO in read-only instances. Fixed the issue where, when multiple tables have complex foreign key dependencies and the foreign key attribute is ON DELETE CASCADE, deleting a parent table record may cause the corresponding child table records to be deleted twice. Fixed the issue where operations such as create user cause a read-only instance to exit. Fixed the issue where executing `show volume status` in a read-only instance may trigger an assertion failure. Fixed the issue where read-only instances return abnormal query results or crash after frequent DDL operations on partition tables. Fixed the crash issue that occurs when a read-only instance query scans a purged undo log, causing a failure during historical version construction. |
2.0.13 | 5.7 | Feature updates Instant add column is supported. For usage, see the Instant DDL feature introduction. Optimizes audit performance under high load and adds the dynamic parameter lock_usleep_time. Issue Fixes Fixed the deadlock issue that could be caused by the official dict_operation_lock during foreign key checks. Fixed the issue where a read-only instance may exit during startup due to replaying acl change logs. Fixed the master-slave data inconsistency issue that occurs after instant add column and truncate table operations. Fixed the issue where log replay errors occur after data is inserted again following instant add column and truncate table operations. Fixed the issue where creating a primary key that includes an instant add column causes an exit. Fixed the issue where read-only instances exit when querying table structures after a partition is created using an instant column or rebuilt. |
2.0.12 | 5.7 | Feature updates Database audit is supported. For usage, see Enable TDSQL-C for MySQL Audit. Supports purge page prefetching to accelerate space reclamation. Updates the table and index size information in system views in real time. Adds additional threads to accelerate the advancement of the recycle lsn and speed up storage garbage collection. Issue Fixes Fixed official full-text index bugs, including: BUG#24938374, BUG#21625016, BUG#27082268, BUG#27155294, BUG#27326796, BUG#27304661, BUG#25289359, BUG#29717909, BUG#30787535. Fixed official bugs related to concurrent updates that could cause system crashes, including BUG#30950714, BUG#31205266, and BUG#25669686. Fixed official bug BUG#28104394, where uncommitted insert operations affect the range scan of index builds, causing them to return incorrect row counts. Fixed official bug BUG#30488700, which addresses the performance issue caused by incorrect derived table query plans. Fixed the issue where a read-only instance may exit after the host executes DDL due to replaying statistics information logs. Fixed the issue where renaming a table to a non-existent database. Fixed the issue where an optimize table operation on the host could cause a read-only instance to exit. Fixed the transaction consistency issue that occurs when full-text index tables are updated on read-only instances. Fixed the deadlock issue that could occur during set offline mode and show variables operations. |
2.0.11 | 5.7 | Feature updates Optimized the binlog file index to improve the speed of scanning binlog files. Optimized the shutdown process to improve shutdown speed. Optimized instance memory by compressing the memory footprint of structures such as buffer, zip, and hash. Optimized the process for recovering DDL locks during read-only instance startup to accelerate log replay speed. Optimized the process for loading system tables on read-only instances to accelerate startup speed. Optimized the allocation of purge coordinator worker threads to improve purge speed. Issue Fixes Fixed the issue where an incorrect index structure mapping caused the system to hang when a table is opened. Fixed the issue where an exception occurred during replication on read-only instances when an xa transaction was rolled back. Fixed the crash issue that occurred during startup when binlog replication is initiated on a read-only instance. Fixed the memory leak issue caused by flush logs. Fixed the issue where the mysqladmin shutdown command fails to exit. Fixed the issue where the read-only instance connection failed to replay logs when the host dropped a column. Fixed the issue where data could still be loaded after the insertion of a large blob object triggered a space limit. Fixed the availability issue with read-only instance replication that may be caused by slow DDL operations on large tables and slow log writing on the host. Fixed the issue of storage waste for small tables after the innodb_max_temp_data_file_size parameter is set for temporary tables. |
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