Community/Database Version | TXSQL Kernel Version | Description |
8.0.30 | 22.7.4 | New features txsql fast path supports partitioned tables. Update the tsm version. TXSQL supports runtime filter. Bug fixes Fix the issue where DECIMAL field BETWEEN range queries incorrectly return empty result sets in const table scenarios. Fix the issue where threads occasionally hang during the group commit process. Port the fixes from TDSQL version 8024 to version 8030. The install_mysql_innodb.sh script supports Chinese locale compatibility. Fix the support for views and CTEs in hyperloglog. Fix the hyperloglog memory leak issue. Fix the issue where Flashback Query returns incomplete historical versions when IN / OR predicates (VERSIONS BETWEEN) are used. Fix the issue where Flashback Query returns incorrect results when ORDER BY is used. Fix the issue where Flashback Query returns incorrect results when VERSIONS BETWEEN with IN/OR conditions on unique keys is used. Optimize error handling and conflict detection for Flashback Query. Fix the issue where Flashback Query: VERSIONS BETWEEN queries cannot return historical versions of records that were inserted and then deleted. Fix the issue where flashback data cannot be retrieved after alter table modify column is executed. Fix the issue where JSON column flashback queries (VERSIONS BETWEEN) return ERROR 3142. Fix the issue where, due to the degradable instance of strong synchronization in kernel 8.0.30, abnormal access occurs for a period after an HA switchover in a primary/secondary node failure scenario, leading to data inconsistency between the primary and secondary nodes. Fix the issue where unsigned BIGINT tables return incorrect empty result sets when compared with negative values. Fix the issue where dropping a database in the recycle bin reports ERROR 1435 (HY000): Trigger in wrong schema. Fix the issue where the decimal operation precision in the kernel is inconsistent with that in SQLEngine. Fix the issue where DDL XA becomes unavailable due to insufficient permissions during connections between CN and DN. Fix the issue where executing DROP TABLE WITHOUT RECYCLE_BIN after a view is created, a table is dropped and restored, and then the table is truncated reports ERROR 1062 (23000): Duplicate entry 'xxx-id' for key 'columns.table_id'. Fix the issue where JOIN queries cause TXSQL Coredump. Fix the issue where CCL rules become ineffective when a connection is made without a default database (thd->db() is NULL). Fix the issue where mysqld crashes after cdb_audit_action_enabled is enabled and CLEAR RECYCLE_BIN <table> is executed when the table name does not include a database prefix. Fix the issue where, when bind_address = * is configured, an error is reported and strong synchronization fails while adding a firewall rule. Fix the issue where, in CCL REFUSE mode under high-concurrency rejection scenarios, show_rule.RUNNING may transiently exceed the CONCURRENCY_COUNT throttling limit. Fix the issue where MDL lock conflicts on intermediate temporary table names during ALTER operations on the secondary node cause the SQL thread to be blocked and primary/secondary synchronization to fall behind indefinitely. Fix the issue where a syntax error (1064) in the DDLXA COMMIT event during replay on the secondary node causes the SQL thread to stop, primary/secondary synchronization to be out of sync, and transactions to be stuck in the prepared state. |
8.0.30 | 22.7.3 | New features The recycle bin feature supports specifying drop time and purge time. The recycle bin's restore database operation supports specifying a timestamp using with TIMESTAMP('xx'). Supports rollback of DML operations within a transaction. A leading hint is appended to the Binlog written by the recycle bin. The recycle bin supports the drop database without recycle_bin operation. Supports obtaining DDL-supported algorithms via native package. Add DDL XA transactions. Supports sharing a readview among multiple statements within the same thread or among statements across different threads. Supports hyperloglog. Adds the performance_schema_session_connect_attrs_size parameter to the configurable parameter list. The query_alloc_block_size&query_prealloc_size parameters are configurable. Data precision fixes based on serialize_data/deserialize_data. Removes the dependency of the recycle bin's RESTORE/CLEAR commands on SUPER privileges. Parameter templates support the block size option. Renames serialize_data to _tdsql_serialize_data and deserialize_data to _tdsql_deserialize_data. Add the txsql_forbid_exists_to_semijoin parameter to the parameter template. Add the tdsql_restrict_database_drop database deletion protection parameter to the parameter template. Bug fixes Fixes the issue where enabling KMS fails on the 8.0 kernel (version 22.6.9 and above) under the TencentOS Server 3.2 (Final) operating system. Fixes the issue where innodb hangs during crash recovery after the thread pool is enabled. Do not use timestamp as the shardkey. Fixes the issue where KMS fails during rollback due to a problem with the kernel pxb package. Fixes MySQL Bug #108981. Fixes the issue where the recycle bin query results are abnormal when indexes are used incorrectly. Fixes the issue of complex sql prepare crashes. Fixes the issue of prepare subquery crashes. Fixes Bug#35531293. Fixes the issue where queries using the WHERE <=> filter condition return incorrect results when comparing against null values. Fixes the issue of crashes caused by complex SQL. Fixes the issue where projection columns return abnormal results in Proxy2.0 + 8030 kernel. |
8.0.30 | 22.7.2 | New features The CREATE TABLE statement cannot specify only the character set without the collation. The jmysql login tool supports the use of hints. Label the Xinchuang edition number and the client prompt. Supports the floor function as a partition key. Configurable parameters for 8030 are output during the packaging process. Remove the connect by syntax in TXSQL mode. The newly added oracle syntax is parsed using a separate syntax tree to prevent the new syntax from affecting the normal mode. Bug fixes Fix the crash issue that occurs in create_buffer_and_save_keyimage. Several community bugs related to AI-based port location have been merged. Specify the auth-plugin for the relevant system account as mysql_native_password. Fix the bug related to binlog_write_threhold. MySQL bug fix: parallel query degradation. Fixes Bug#35039937: Multi-value virtual columns should not be materialized. Fix the issue where the thread count for Parallel_reader poses a risk of memory leakage. Fix the issue where update histogram does not report an error when it fails to obtain a thread. Fix the issue where the cardinality estimation for non-partitioned columns in partitioned tables has a large error. Fix the issue where repairing the record of a node's first xa transaction may cause other select withgts statements to see inconsistent views. Use the partitioned udf for post-processing to handle the coredump caused by part_key being null. Fix the issue where the tdsql_oracle_mode_ignored_users parameter does not take effect after a restart. Add the noop_no_mono function to partition_udf. In oracle mode, the results are returned in uppercase. Fixes the issue where the hot update feature, which failed during long-term stability recovery, causes continuous restarts. Fix the issue where, under concurrent load, the rec_per_key for partitioned tables may become infinite (inf), causing a crash. |
8.0.30 | 22.7.1 | New features Optimize the performance of Prepared statements or Stored Procedures by caching the results of records_in_range. Optimize the performance of Distinct in Sysbench scenarios. Change the version number to 8.0.99-v24-txsql-22.7.0-20240718 and keep innodb_version as 8.0.30. Modify the tools that depend on libssl/libcrypto to be statically compiled. Supports displaying the KMS plugin version by executing the command `show global status like 'keyring_kms_version'`. Supports saving at least the two most recent encryption configurations and master keys to ensure fault recovery. Provides a tool to determine whether the keyring file can decrypt the binlog. Add the txsql_forbid_normal_user_login parameter, which, when enabled, prevents regular users from logging in. Supports dble custom partitioning algorithms and enables pre-installation startup via plugins. Adds format validation to the tdsql_sub_version parameter. Connections to TDSQL compute nodes ignore the returned affected row when CLIENT_FOUND_ROWS is set. Adds the tdsql_connection_id variable and enables rand_seed, last_insert_id, and tdsql_connection_id to be updated by the push-down statement HINT. Connections to TDSQL compute nodes bypass the NO_ZERO_DATE check to allow the execution of zero datetime literals. Supports the oracle lengthb function. Supports the sum, avg, and count functions returning the Decimal type in Oracle mode. Supports the NVL2 function. Adds a switch to control the case of column headers in Oracle Mode. Bug fixes Fixes the performance regression issue caused by unnecessary os_file_flush and sets the default values of the resource_group and plan_cache system variables to false. Fixes the issue where DML operations might coredump because trx_sys->shards were not properly cleaned up during crash recovery. Fixes the issue where executing SQL statements that do not require parsing (such as quit) might cause a coredump when tdsql_log_slow_extra is enabled to output more information in the slow log. Fixes the issue where a deadlock might occur when statistics histogram and alter table operations are executed concurrently. Fixes the issue where a syntax error is reported when an insert statement contains advisor/density as a column name. Fixes the issue where incorrect null value handling in the murmurHashCodeAndMod_CL partitioning function might cause a coredump. Fixes the Unicode mapping error issue for GB18030 characters. Fixes the issue where, in strong synchronization mode, a transaction that kills a query/connection must wait for the ack to be committed. Fixes the issue where modifying sqlasync_wait_slave_hosts might cause strong synchronization to be stuck indefinitely. Fixes the issue where, in audit logs, multiple SQL statements executed consecutively by the same thread within a very short time might appear to have the same execution timestamp. Fixes the issue where the execution timestamp in audit logs is not updated after cdb_enable_resource_statistics is enabled. Fixes the issue where data is not accurately filtered when a query is made with a string column condition value that exceeds the column width after txsql_fast_path_mode is enabled. Fixes the issue where a query using a non-unique secondary index might return fewer data rows after txsql_fast_path_random_mode is enabled. Fix the issue where the cardinality estimation for non-partitioned columns in partitioned tables has a large error. Fixes the issue where instance creation fails due to the lack of the frominstall_my_port.ini configuration. Fixes the issue where the Partition Description column in the information_schema.PARTITIONS table for secondary partitions contains incorrect values. Fixes the UNION error caused by the NULL character set in Oracle mode. |
8.0.30 | 22.7.0 | New features Provides the orzdba tool and perl monitoring scripts. Auditing supports recording transaction IDs and xids. Adds profile to the slow log. sqlasync_wait_slave_hosts supports specifying a slave by using the server_id. MySQL supports establishing primary/secondary replication relationships by using domain names. The slow log supports recording mutex wait time. Update kernel security-related components. Optimizes the impact of xa commit on writeset. Supports the port fast path feature. Supports columns in subpartition. Supports creating default list partitioned tables. Supports using UDFs in first-level list/range partition expressions. Supports semi-join/anti-join syntax. Optimize the performance of ha_recover. Supports hierarchical queries using connect by. Bug fixes Fix the issue where blob/longtext data types cause table data to be unreadable and unwritable. Fix the issue where the kernel incorrectly marks a transaction as an XA transaction when a non-XA transaction request is made after an XA transaction has been executed. After LWG-3865 is fixed, the equality operator needs to be updated. Fix the issue where enabling the recycle bin causes a coredump when lock table or truncate operations are executed. Fix the issue where the mysql client becomes unavailable due to arm compilation optimization. Fix the issue where the initialization of the Executed_Gtid_Set on the primary node fails abnormally after a restart when the primary's gtid is less than that of the secondary. Fix the issue where virtual columns in secondary partitions cause abnormal queries. Remove unnecessary files from the TencentSM component. Fix the issue where, during a DELETE operation, the filter is ignored when a materialized table exists in a subquery, causing the condition to not take effect. Fix the issue where the database crashes after a DELETE / UPDATE statement that contains a subquery is executed. Fix the issue where executing SQL results in the error "Unknown column 'copy_range_range_1.col_decimal' in 'having clause'". Fix the issue of inconsistent transaction data for transfers in the fault-tolerant long-term stability (noshard) table. Fix the issue in consistent read: when min_purge_gts is 0, the system still performs purge using the single-node logic even if mc_enable is enabled. Fix the issue where the anti join syntax causes a crash. Fix the issue where, in Oracle mode, the query SELECT TO_NUMBER('3.33','') causes the sqlengine to exit abnormally. Fix the issue of incorrect calculation results for bit-type aggregation. Fixed the issue where supporting the operation `alter table test_t.sbtest3 drop subpartition template p_20220514` triggers a crash. Fix the issue of add partition rollback crashes. Fix the performance degradation issue of where in statements after prepared statements are used. Fix the issue of duplicate key errors occurring when global indexes are used. Fix the issue where creating a table using `create TABLE ... like` in a secondary partition causes a crash. Fix the issue where the assertion `m_prebuilt->partkey_templ_pos_array_size() == partkey_num` fails for global indexes. Fix the issue of enqueue/dequeue inconsistency in the global index prefetch cache. Fix the issue where txsql_returning causes a crash. Fix the issue where using an uninitialized length array during column pruning causes a crash in the release version. Fix the issue of data query errors in secondary range columns partitions. Fix the issue of incorrect result sets returned by the MAX function. Fix the issue where the presence of constant functions in `show create view` causes a crash. Fix the issue where missing collate information during murmurhash calculation causes a crash when an insertion into a global index is performed. Fix the issue where the application order of online logs is inconsistent with the data write order. Fix the issue of unreleased memory discovered during the code scanning process. Fix the issue where using uncommitted indexes for shared access causes a crash. Fix the issue where TRUNCATE SUBPARTITION TEMPLATE causes a crash. Fix the issue where creating a subpartitioned table without specifying a subpartition template triggers a crash. Fix the issue where a crash is triggered when an uninitialized subpartition template is checked. Fix the issue where concurrent transactions acquire excessive locks on secondary indexes. Fix the issue where tlog files are not cleaned up after the disaster recovery DCN disables the MC. Fix the issue where the standby database hangs while waiting for a Semaphore. Fix the issue where using distinct + with rollup in TXSQL results in missing data. Fix the issue where the keyring_kms plugin causes mysqlbinlog to crash when the encrypt-key-file size is zero. Fix the issue where the Server crashes in the intern_plugin_lock function during concurrent plugin installation. Fix the issue where the performance of sqlasync_group_slave_ack drops to zero when strong synchronization is disabled. Fix the issue where performance drops to only 10% of the original after binlog compression is enabled. Fix the issue where inserting data into a table containing an auto-increment column causes a crash, and the crash recovery process upon restart consumes a large amount of memory. Fix the issue where the database fails to start after the federated parameter is added. Fix the issue where the system crashes after derived_condition_pushdown is enabled. Fix the issue where a crash is triggered when an Instant column is added to a primary key. Fix the issue where pxb, when operating on xfs, causes abnormal copying of compressed data, leading to space inflation. Fix the issue where performing a truncate operation on a global index table concurrently with DML operations may trigger a database coredump. Fix the issue where a crash is triggered when CHECK PARTITION ALL is executed on a global index. Fix the issue where a crash is triggered when global index data is released. Fix the issue where executing alter TABLE truncate partition on a global index with an empty table instance causes a crash. Fix the issue where a crash is triggered in the subpartition expression error handling path. Fix the issue where a purge operation on a global index may trigger a crash. Fix the issue where a crash is triggered when RANGE COLUMNS subpartitioning is used and multiple identical columns are specified. Fix the issue where MODE_ORACLE-related assertion failures are triggered in binlog scenarios during MTR tests. Fix the issue of semi join crashes. Fix the issue where truncate causes a database crash. Fix the issue where transaction visibility problems may occur in the xa commit one phase scenario. Add a global parameter to dynamically disable queries to the data_locks, data_lock_waits, and innodb_trx tables. Fix the issue where the partition_description column in information_schema has an incorrect format when using list columns with a default partition. Fix the issue where queries on partitioned tables under the gb18030 character set report errors. Support returning field names in uppercase in Oracle mode. The innodb_async_drop_tmp_dir parameter has been added to the parameter template. Fix the issue where TXSQL handles the level keyword abnormally in connect by statements. Remove support for the CONNECT BY syntax in TXSQL mode. Decouple Oracle syntax from MySQL syntax. Fix the issue where the REORGANIZE PARTITION syntax does not take effect. Fix the issue where the m_udf_state_mutex for temp tables is not properly destroyed. Port the fix for Bug#35289390: purge_sys->thds.find(thd) has performance hotspots / lock contention. Fix the issue where an excessive number of columns causes unpack to fail during binlog replay. Fix the issue where sending SQL statements without GTS via passthrough statements under the condition of enabling consistent reads can cause visibility judgment errors. Fix the issue where a single modification statement in audit logs lacks a transaction id. Fix the parameter type of txsql_parallel_factor to use the ulong type. sqlasync_wait_slave_hosts supports specifying a slave by using the server_id. Fix the issue where the system crashes after derived_condition_pushdown is enabled. Fix the issue where executing a custom decode function in normal mode reports the error "func decode is not support in none oracle mode". Fix the error message for rejecting a multi-table delete statement with a returning clause. Remove the ulimit settings from the TDSQL bootmysql startup script. Fix the issue where the truncate operation on a secondary partition does not take effect after a column is deleted from the table. Fix the issue of incorrect decimal results. Fix the issue of incorrect calculation for the year data type. Fix the issue of incorrect results for Left Scalar Join. Fix the issue where a database coredump occurs when the check partition all operation is performed during DML operations on a secondary partitioned table that contains a global unique index. Change txsql_subpartition_support_multiple_columns to be enabled by default. Fix the issue where the query results for cte queries differ between version 8030 and version 8024. Fix the issue where an error is reported when related operations are performed after txsql_oracle_mode=on is enabled and tables/functions containing semi/anti fields are created. Fix the issue where partitions are not correctly pruned by Range columns. Fix the issue where statistical information is lost after the column length is modified (rebuild table). Fix the issue of unique key conflicts occurring when global indexes are used. The parameter performance_schema_show_processlist is set to on in the factory configuration file. Fix the issue where an error is reported when a single partition is truncated. Fix the issue where an exception occurs when a check index operation is performed on a full-text index. Fix the issue where SHOW CREATE TABLE exhibits abnormal behavior when DEFAULT values for non-UTF8 character sets are displayed. Fix the issue where the txsql_check validation rule is lost after CREATE TABLE LIKE is used. Fix the issue where the display in txsql_check is inconsistent before and after an alter table operation. Modify the processing collation for the expression in txsql check to my_charset_bin. Fix the issue of id overflow in partitioned tables. Fix the issue where sending kill -1 / kill -10 signals causes mysqld to restart. Fix the issue of potentially abnormal ptr for host_or_ip in mpvio. Fix the issue where PS fails to correctly select an index when the index condition contains a parameterized subquery. Fix the issue of abnormal statistics for the Innodb_data_pending_fsyncs metric in mysql 8.0. Extends slowlog information with profile timing data. Fix the issue of abnormal query time in slow queries. Fix the issue where the TDSQL client does not disconnect upon timeout. Fix the issue of Encountering the assertion failure "'max_length == 4' failed for handle cast year in tmp_table_field_from_field_type". Fix the issue of incorrect logic in subpartition pruning. Fix the issue in RANGE/LIST subpartition testing. Fix the issue where the ADD PARTITION operation fails on secondary partitioned tables. Fix the issue of duplicate key errors occurring in scenarios with concurrent DROP PARTITION and DML operations (the specific key value is ???).Fix the issue where OOM occurs when the dd::client loads a large number of partitioned tables containing auto-increment columns and starts the Server. Fix the issue of assertion failure in txsql_returning. Fix the issue where errors occur in DML when column names are semi/anti keywords or function calls. Fix the issue where insert/update statements that use the same table cache fail after a load DATA operation imports empty data into a non-nullable field. Fix the issue of incorrect result sets generated by start with connect BY. Modify the output format of xid in audit logs. |
8.0.30 | 22.6.8 | Bug fixes Fix the issue of data loss during frequent rebuild table operations. pxb supports global indexes and columns. |
8.0.30 | 22.6.7 | Bug fixes Fix the issue where transactions become invisible after an xa commit. Fix the issue where txsql_disable_gtid_in_undo is disabled in the parameter template. Fix the issue where enabling the recycle bin causes a coredump when lock table or truncate operations are executed. |
8.0.30 | 22.6.2 | New features The features of TDSQL 8030 are aligned with those of MySQL 8030. Bug fixes Fix the issue where the mysqld.err file contains excessive log entries. |
Community/Database Version | TXSQL Kernel Version | Description |
8.0.24 | 22.6.11 | Bug fixes Fix the issue where modifying the version number causes prepared statements to fail to retrieve results. Fix the issue where the murmurHashCodeAndMod_CL function does not handle null_value when processing string-type data. |
8.0.24 | 22.6.10 | Bug fixes Fix the issue with the community MySQL CPU limit. |
8.0.24 | 22.6.9 | New features A new status variable has been added to display the KMS plugin version. A new kernel tool has been added to determine whether the binlog can be decrypted using the keyring file. The kernel supports saving at least the two most recent encryption configurations and master keys to ensure fault recovery. SQL Filter is supported. The slow log supports the optimizer trace feature. Upgrade the dependency libraries used by TXSQL for the community version 8.0.41. Global index information can be collected via dictionary tables. Bug fixes Fix the issue where the binlog parser in version 8.0.24 is incompatible with the xa prepare event in the binlog of version 8.0.18. Fix the issue where, when mc is enabled, multiple restarts before commit after a one-phase transaction completes prepare may cause the prepare gts to be mistaken for the commit gts, potentially leading to partial reads of the transaction. Fix the issue where enabling PAD_CHAR_TO_FULL_LENGTH in sql_mode causes the replication process to become abnormal (official bug, requiring a forced upgrade with --upgrade=force to succeed). Fix the issue where, in a stored procedure/function, nested calls to the sm4 decryption algorithm cause all subsequent calls to fail after a single failure occurs. Fix the issue where the database experiences a coredump after the parameter `set global log_slow_verbosity = 'explain';` is configured. Fix the issue where modifying sqlsaync_wait_slave_hosts may cause strong synchronization to be stuck indefinitely. Fix the issue where the strong synchronization mechanism may not take effect after reset master. Fix the issue where enabling oracle mode causes the redo operation on the standby server to fail. Fix the issue of high master-slave latency in groupshard. Fix the issue where the kernel assertion fails because the reference count bug_fix_count for the data page is not zero. Fixed the issue where the purge thread incorrectly clears undo when min_purge_gts = 0 and mc_enabled = ON. Fix the issue where cte query results are incorrect. Fix the issue of inaccurate trx_started in innodb_trx. Fix the issue where "Open file list xxx" is printed multiple times in the error log. Fix the issue of abnormal statistics for set-level global index space. |
8.0.24 | 22.6.8 | New features Supports restricting user connection access in security mode. Bug fixes Fix the issue where in-type SQL executes significantly slower than or/and-type SQL when there are many values. Fix the issue where SQL queries with a range of Item_row are not optimized. Fix the issue where enabling the derived_merge parameter may cause the optimizer to convert a left join to an inner join during SQL rewriting, resulting in incorrect results. Fix the issue of incorrect query results with NOT IN statements. Fix the issue where stats_auto_recalc times out while obtaining MDL during DDL execution, causing statistics for prefix partitions of partitioned tables not to be collected. Fix the issue where transaction rollback involving TRX_UNDO_UPD_DEL_REC type undo may cause inconsistent query results in global consistent reads. Fix the issue where trx_i_s_possibly_fetch_data_into_cache accesses memory that has already been freed. Fix the issue where a segfault and crash occur when to_char is called with a NULL date value and an ER_TRUNCATED_WRONG_VALUE error is present during execution. Fix the issue where killing a connection causes a transaction to be committed in strong synchronization after sync mode. Fix the crash issue caused by the show slave uuid command. Fix the issue where, if a materialized table exists in a subquery, the filter is ignored, causing the condition to not take effect and resulting in unstable query results. Fix the issue where executing SQL requests with the WHERE clause (v = (select ?)) using the jdbc prepared statement method causes a crash. Fix the issue where, during the XA COMMIT phase, if writing the gts to redo fails, the transaction is committed with an incorrect gts. Fix the issue where complex SQL causes a kernel crash. Fix the issue of inconsistent results from complex query unions. Fix the issue of incorrect results for in subqueries under the firstmatch policy. Fix the crash issue caused by merging derived tables and window functions. Fix the issue where killed transactions cannot be automatically committed via the ack of the high-watermark binlog. Enhance the error message for insert operations on secondary partitioned tables. Fix the issue where the murmurHashCodeAndMod function is case-sensitive in certain CREATE TABLE statements. |
8.0.24 | 22.6.7 | New features The slow log supports recording mutex wait time. Add the txsql_forbid_query_internal_table parameter to dynamically disable queries to the data_locks, data_lock_waits, and innodb_trx tables. Supports in-place upgrade from version 8024 to version 8030. Supports the FEDERATED engine. Bug fixes Fix the crash issue that occurs when duplicate columns appear in a subpartition by range columns clause during table creation while the txsql_subpartition_support_multiple_columns parameter is enabled. Fix the issue where the performance of sqlasync_group_slave_ack drops to zero when strong synchronization is disabled. Fix TCE-related vulnerabilities. Fix the issue of add partition rollback crashes. Fix the issue where subsequent insert/update statements fail after a load DATA operation imports empty data into a non-nullable field. Fix the issue where performance drops to only 10% of the original after binlog compression is enabled in a strong synchronization scenario. Fix the issue where the initialization of the Executed_Gtid_Set on the primary node fails abnormally after a restart when the primary's gtid is less than that of the secondary. Fix the issue where tlog files are still generated when the MC is disabled. Fix the issue where the system crashes after derived_condition_pushdown is enabled. Fix the issue where enabling the recycle bin causes a coredump when lock table or truncate operations are executed. Fix the issue where the check index operation crashes in a full-text index scenario. Fix the issue of PSI_Mutex memory leakage in the memory table. Fix the issue of semi join crashes. Fix the issue where the wrong index is selected for equality queries when the index cardinality is high (close to that of the primary key), due to the higher cost of using a global index compared to a local index. Fix the issue where an error is reported when subpartitions are truncated if the number of selected partitions or subpartitions is greater than or equal to the total number of partitions or subpartitions in the current table. Fix the issue where errors are reported in dml when column names are semi/anti keywords or function call keywords. Fix the issue where a prepared statement containing a subquery with all parameters leads to slow execution due to a full table scan. Fix the issue where the returning statement causes a crash. Fix the issue of incorrect result sets returned by the max function. Fix the issue where blob/longtext data types cause table data to be unreadable and unwritable. Fix the issue of abnormal statistics for the Innodb_data_pending_fsyncs metric. Fix the performance degradation issue of where in statements after prepared statements are used. Fix the issue where creating a table using `create TABLE ... like` in a secondary partition causes a crash. Fix the issue of abnormal query time in slow queries. Fix the issue where the strong synchronization status waiting_thread_count is displayed as a negative number. Fix the issue of incorrect results from bit-type aggregation calculations. Fix the crash issue caused by constant functions in `show create view`. Fix the issue of a potentially abnormal pointer for host_or_ip in mpvio. Fix the issue where `group by`/`distinct` operations do not meet collation expectations when strings contain trailing spaces. Fix the issue where the tdsql_oracle_mode_ignored_users parameter does not take effect after a restart. Fix the issue with the level keyword in connect by statements. Fix vulnerability CVE-2024-21015. |
8.0.24 | 22.6.6 | Bug fixes Fix the issue where the system crashes after derived_condition_pushdown is enabled. Fix the null pointer issue caused by the udf initialization process incorrectly entering the oracle connect BY function logic. Fix the issue where a segfault and crash occur when to_char is called with a NULL date value and an ER_TRUNCATED_WRONG_VALUE error is present during execution. |
8.0.24 | 22.6.5 | Bug fixes Fix the issue where the unique index cannot be created successfully due to the inconsistency between the Replay order and the actual write order of the online log in the unique global index. Fix the issue where the global index drop subpartition template crashes during rollback. Fix the issue where a crash occurs when rebuilding the primary key after an instant column is added. Fixed the issue of a crash caused by using uncommitted global indexes when a table is opened. Fix the issue where copying an uninitialized partition key field in the prebuilt->mysql_template within a global index table causes an out-of-bounds memory write, resulting in a crash. Fix the issue of an out-of-bounds read occurring during the subpartition template copy process when a partition is added. Fix the issue where the error record is reported as ??? when a duplicate record rollback occurs during global unique index DDL.Fixed the issue of the memory leakage during global index rollback. Fix the issue where a crash occurs when a COLUMNS partitioned table is created with duplicate columns in the partition key. Fix the issue where an error is reported when a subpartition is truncated while there is only one partition. Fix the issue where virtual columns in secondary partitions cause abnormal queries. |
8.0.24 | 22.6.4 | Bug fixes Fix the issue where, due to xfs pre-allocating a significantly higher number of blocks than actually needed when writing to files with tail holes, the data disk utilization in the disaster recovery environment becomes much higher than in production. This may cause the data disk to become full during dcn setup or during the process of rebuilding a standby server. |
8.0.24 | 22.6.3 | Bug fixes Fixed the issue where concurrent DML and DDL operations in global consistent read mode cause inconsistencies between the innodb layer data dictionary and the table cache. Fix the issue where a table containing global indexes only reports an error but does not crash when metadata is corrupted. Fix the issue where, under the Read Committed (RC) isolation level, a transaction performing an insert + update/delete can have its uncommitted insert record seen by another transaction. Fix the crash issue caused by global index tables under concurrent inplace + rebuild DDL and DML operations. Fix the memory leak issue in temporary tables. |
8.0.24 | 22.6.2 | Bug fixes Fix the issue where, when the global consistent read RC isolation level is enabled, XA ONE PHASE and autocommit single-statement transactions may experience read inconsistency. |
8.0.24 | 22.6.1 | New features sqlasync_wait_slave_hosts supports specifying a slave by using the server_id. Supports hierarchical queries using connect by. Supports using UDFs in first-level list/range partition expressions. Supports semi-join/anti-join syntax. Supports creating default list partitioned tables. Bug fixes Fix the issue where, due to the lack of collate information for characters in murmurhash, duplicate records may be inserted into the global index, subsequently triggering an assertion and causing a crash. Fix the crash issue related to the global index that occurs when the access key in a SELECT statement is the primary key and the SELECT list does not include all partition keys. Fix the issue where TRUNCATE SUBPARTITION TEMPLATE causes a crash. Fix the issue of duplicate key errors when global indexes are used. Fix the issue where statistical information is lost after column length is modified (rebuild table). Kernel default parameter template optimization: Enable checksum. txsql_enable_resource_statistics is disabled by default. Fix the issue of data query errors in secondary range columns partitions. Fix the issue where executing a custom decode function in normal mode reports the error "func decode is not support in none oracle mode". Fix the issue where the REORGANIZE PARTITION syntax does not take effect on secondary partitioned tables. |
8.0.24 | 22.6.0 | New features Strong synchronization supports establishing primary/secondary replication relationships by using domain names. Supports the extension of slow log information. Supports the partition exchange logic based on secondary partitions. Audit logs support recording internal_xid and trx_id. Bug fixes Fix the crash issue caused by redo corruption when space is full. Fix the issue where the keyring_kms plugin causes mysqlbinlog to crash when the encrypt-key-file size is zero. Fix the issue of a nullptr exception when a schema table is referenced by internal queries. Fix the issue where a purge operation may trigger a crash in global indexes. Fix the issue where executing alter TABLE truncate partition on an empty table instance within a global index causes a crash. Fix the issue where the standby database hangs while waiting for a Semaphore. Fix the issue where the data dictionary causes an OOM during startup by loading a large number of partitioned tables containing auto-increment columns. Fix the issue of assertion failure in txsql_returning. Fix the issue of a crash occurring during sqlsmith testing within replace_columns_in_cond. Fix the issue where the kernel incorrectly marks a transaction as an XA transaction when a non-XA transaction request is made after an XA transaction has been executed. Fix the issue where unpack fails during binlog replay due to an excessive number of columns. Fix the issue of abnormal output when show create table default value is executed on non-UTF8 character sets. Fix the issue where sending SQL statements without GTS via passthrough statements when consistent reads are enabled can cause visibility judgment errors. Fix the issue where using distinct + with rollup in TXSQL results in missing data. Fix the issue in range/list subpartition testing. Fixed the issue where InnoDB Instant DDL (dropping and then adding a column with the same name) causes table metadata corruption and crashes. Fix the issue where the install_mysql_innodb.sh script requires cnf_file_tmp to be chowned to tdsql:users. Fix the issue where concurrent transactions acquire excessive locks on secondary indexes. |
Community/Database Version | TXSQL Kernel Version | Description |
8.0.18 | 22.6.1 | Performance optimization Kernel default parameter template optimization: Enable checksum. txsql_enable_resource_statistics is disabled by default. |
8.0.18 | 22.6.0 | Bug fixes Fix the issue of slow performance for instant ddl add column operations. Fix the issue where, when permissions are granted, the records in the binlog do not meet expectations, causing slaves and RO nodes to fail to obtain the correct permissions, which may lead to a large number of business errors after a switchover. |
8.0.18 | 22.4.0 | New features The install_mysql_innodb.sh/mysql-config.py scripts are optimized to support setting the directory permission group to the username, instead of 'users'. Bug fixes Fix the issue where the analyze table operation causes a large number of slow queries to accumulate in the 'Waiting for table flush' state. The deadlock detection mechanism in multi-threaded replication is modified, support for MDL/ACL locks is added, and the issue where deadlocks might occur in MTS+slave_preserve_commit_order mode is fixed. Fix the crash issue caused by the page cleaner's assertion that sleep > 0. |
8.0.18 | 22.3.0 | Bug fixes Fix the issue where special characters in auditing are not properly escaped, causing single and double quotes to display abnormally. Fix the issue where setting skip_scan=ON causes incorrect data to be read. Fix the issue where adding a LONGTEXT column via inplace ddl may cause a crash. Fix the abnormal issue with updating JSON field composite indexes. Fix the issue of server-id conflicts. Fix the issue where the install_mysql_innodb.sh script incorrectly determines the aio logic. Fix the abnormal usage issue caused by the event_scheduler system user not being on the allowlist. |
Community/Database Version | TXSQL Kernel Version | Description |
5.7.17 | 22.6.13 | Bug fixes Fix the issue where instance provisioning fails due to some tools bundled with the Percona 5.7.17 kernel relying on OpenSSL 1.0. Fixes BUG#99286. Fix the issue where a move-back tlog table is incorrectly treated as a tlog file, causing rollback failure. |
5.7.17 | 22.6.8 | New features Adds the tdsql_forbid_write switch. Bug fixes Fix the issue where slow queries are caused by incorrect statistics of the thread pool connection_count. |
5.7.17 | 22.6.7 | Bug fixes Fix the issue of abnormal conversion of m_is_mysql_user_table. Update the mysqld version to 5.7.36. |
5.7.17 | 22.6.1 | New features Supports disabling the log_prepared_xid_list. |
5.7.17 | 22.6.0 | Bug fixes Fix the issue where the server crashes when querying partitioned tables. |
5.7.17 | 22.5.0 | New features Supports the recycle bin. Bug fixes Fix the issue where alter table causes a crash. Fix the crash issue caused by incorrect lock ordering. Fix the issue where sending kill -1 / kill -10 signals causes mysqld to restart. Fix the issue where the show grants command displays abnormal results. Fix the issue where the Server crashes in the intern_plugin_lock function during concurrent plugin installation. Fix the issue where permission checks for views containing subqueries are abnormal. |
5.7.17 | 22.4.0 | New features The install_mysql_innodb.sh/mysql-config.py scripts are optimized to support setting the directory permission group to the username, instead of 'users'. Performance optimization Enhance the binlog decoding process's ability to handle binlogs that are accidentally corrupted. Bug fixes Fix the issue where the CASE, COALESCE, IF, and IFNULL functions use an incorrect data type when merging signed and unsigned integers. Fix the issue where a corrupted USER table at startup does not generate an Error Log. Fix vulnerability CVE-2017-3599. Fix the issue where the optimizer may return incorrect result sets when using the index_merge_intersection method. |
5.7.17 | 22.3.0 | Bug fixes Fix the issue of abnormal exec_time in the binlog sequence. Fix the server-id conflict issue. Fix the abnormal issue of high CPU load under no-load conditions during serial replication. Fix the potential OOM issue that may occur when PXB fails to execute the flush table with READ lock command. Fix the issue where killing a transaction with a rollback savepoint results in data being committed without being recorded in the binlog. Fix the issue where the install_mysql_innodb.sh script incorrectly determines the aio logic. Fix the abnormal usage issue caused by the event_scheduler system user not being on the allowlist. |
フィードバック