polardbx-sql-5.4.15-16792770

RELEASE NOTE

Feature enhancement

Add support for information_schema privileges system tables compatible with MySQL.
Add support for viewing space usage of archive tables.
Add support for moving partitions of original tables when binding with archive tables.
Add support for partition hint.
Add support for high privilege users to kill other connections.
Add support for Binary and VarBinary types as partition keys in AUTO mode databases.
Optimize logic for creating time column indexes for auto partition tables.
Optimize private protocol probing process.
Optimize query logic with AGG functions under TSO transactions.
Optimize global index selection logic.
Optimize performance under JDBC protocol.
Optimize speed of row count collection.

Bugfix

Fix the issue where some views queries on information_schema resulted in errors.
Fix the issue where load data statements resulted in errors when specifying character sets without quotation marks.
Fix the issue where mock mode did not support UPSERT.
Fix the issue where unique global indexes may have inconsistent data with the main table during creation or data migration.
Fix the issue where single-point queries on partition keys could result in incorrect query results due to parallel writes in AUTO mode databases.
Fix the issue where altering the AUTO_INCREMENT property of a single table in DRDS mode databases automatically associated with Sequence.
Fix the issue where type matching errors occurred in UPSERT when the Update list order differed from table definitions.
Fix the issue where no error message was displayed when Kill Connection failed.
Fix the issue where information_schema permission verification failed multiple times in prepare mode queries.
Fix the issue where distributed transaction recovery and excessive suspended transactions may lead to DN deadlocks when DN sessions are full.
Fix the issue where uuid and uuid_short functions were calculated only once.
Fix the issue where unexpected errors or legacy DDL tasks happen in some timeout or interruption scenarios.
Fix the issue where pruning led to excessive CPU usage.
Fix the issue where analyze table statements caused CN out of memory.
Fix the issue where spm params cleaning failed under high concurrency.
Fix the issue where AGG functions failed because SQL_SELECT_LIMIT sent to DN.
Fix the issue where the main index ndv was not processed during analyze of tables.
Fix the issue where physical tables could not be found after renaming.
Fix the issue where can not update rows with a primary key value of "0000-00-0000:00:00" in JDBC protocol.
Fix the issue where sample tables were not synced in the scheduled statistics information task.
Fix the issue where expired logic judgment bugs caused sample tasks to fail to be initiated.
Fix the issue where private protocol network layer was unexpectedly disconnected in poor network environments.
Fix the issue where MPP query mode on read-only instances was prone to deserialization failures.
Fix the issue where SQL_SELECT_LIMIT set to default resulted in errors.
Fix the issue where <=> rcausing partition pruning to be ineffective and requiring full table scanning.
Fix the issue where CURRENT_TIMESTAMP was not updated in specific UPSERT scenarios.
Fix the issue where errors occurred in some Semi-join scenarios with GSI.
Fix the issue where some DAL statements had incorrect table-level permissions.
Fix the issue where syntax errors occurred when host was empty in set password statements.
Fix the NPE issue in RawString when receiving null values.
Fix the issue where alter table remove partitioning failed when the primary key and sharding key were the same.
Fix the issue where unexpected primary key conflicts during data migration could slow down the process.
Fix the issue where large-scale DN nodes may stall and fail when HA is enabled simultaneously.
Fix the issue with decimal value processing during the execution of the sum aggregate function.
Fix the issue where the ENABLE_BACKGROUND_STATISTIC_COLLECTION configuration did not take effect.
Fix the issue where implicit primary keys were exposed during show full columns, causing sharding JDBC to fail.
Fix the issue where errors occurred when executing REPLACE and UPSERT commands on tables with JSON-type columns.
Fix the issue where the found_rows function did not calculate the correct result in stored procedures.
Fix the issue where the topN was lost in statistics information when column names were in uppercase.
Fix the issue where the DDL task may fail and cannot continue or be rolled back when rebuilding composite primary keys and deleting some primary key columns with alter table commands.
Fix the issue where the json_extract function incorrectly returned the string "NULL".
Fix the issue where alter table drop column may result in the inability to access the table.
Fix the issue where Lookup Key in BkaJoin with GSI in AUTO mode databases was inferred incorrectly.
Fix the issue where private protocol displayed error messages when there were insufficient backend connections.
Fix the issue where private protocol may deadlock in high-load CN.
Fix the issue where archive table queries may occasionally hang.
Fix the issue where probes in private protocols may terminate connections due to large packet transfers.
Fix the issue where multi-table joins in DRDS mode has incorrect execution plans.
Fix the issue where TTL tables became invalid after RENAME.
Fix the issue where implicit type conversion processing errors occurred in varchar-type partition keys when writing negative numbers.
Fix the issue where executing materialized subqueries with an estimated number of rows of 500 may result in incorrect execution plan construction.
Fix the issue where private protocol probing may cause connection shaking due to a 10-minute delay.
Fix the issue where permission verification failed when analyzing tables across databases.
Disable alter table convert to character set.
For auto partition tables, the default partition number for their global secondary indexes has been changed from instance level to table level.

发布说明

特性更新

新增 information_schema 权限系统表兼容 MySQL。
新增支持查看归档表占用空间。
新增支持在绑定归档表的情况下对原表进行 move partition。
新增 partition hint 功能。
新增支持高权限用户 kill 其他连接。
新增 AUTO 模式数据库中分区键支持 Binary 和 VarBinary 类型。
优化默认主键拆分表建立时间列索引的逻辑。
优化私有协议探活流程。
优化 TSO 事务下带 AGG 函数的查询逻辑。
优化全局索引的选择逻辑。
优化 JDBC 模式下的性能。
优化 row count 采集的速度。

问题修复

修复 information_schema 的部分视图查询报错的问题。
修复 load data 语句指定字符集时不加引号报错的问题。
修复 mock 模式不支持 UPSERT 的问题。
修复 unique global index 在创建或者数据迁移过程中，可能出现与主表数据不一致的问题。
修复分区键单点查询有概率因 Auto 模式数据库写并行导致查询结果不正确的问题。
修复 DRDS 模式数据库中通过 ALTER TABLE 变更单表的列 AUTO_INCREMENT 属性时，自动关联 Sequence 的问题。
修复 UPSERT 中 Update list 顺序与表定义不同时，类型匹配报错的问题。
修复 Kill Connection 失败时没有任何报错的问题。
修复 prepare 模式下，多次查询 information_schema 权限校验失败的错误。
修复分布式事务恢复、悬挂事务过多且 DN 会话打满时可能导致 DN 死锁的问题。
修复 uuid 和 uuid_short 函数仅计算一次的问题。
修复在一些超时或中断场景下非预期报错或遗留 DDL 任务的问题。
修复 pruning 导致占用过多 CPU 的问题。
修复执行 analyze table 语句时计算节点内存超载的问题。
修复 spm params 高并发下的清理问题。
修复 SQL_SELECT_LIMIT 下发至数据节点引发的 AGG 函数正确性问题。
修复统计信息在某表没有条件访问时，主动 analyze 没有处理索引 ndv 的问题。
修复 rename 之后找不到物理表的问题。
修复包含 GSI 的表上无法通过 JDBC 更新主键为”0000-00-00 00:00:00“值的数据行的问题。
修复统计信息定时任务 sample table 没有 sync 的问题。
修复过期逻辑判断 bug 导致 sample 任务无法发起的问题。
修复网络环境较差时，私有协议网络协议层非预期断开的问题。
修复 MPP 查询模式在只读实例上易出现反序列化失败的问题。
修复 SQL_SELECT_LIMIT 设置为 default 时报错的问题。
修复 <=> 导致 分片裁剪失效而采用全分片扫描的问题。
修复特定情况下 UPSERT 没有更新 CURRENT_TIMESTAMP 的问题。
修复带 GSI 的部分 Semi-join 场景产生报错的问题。
修复部分 DAL 语句对应表级别权限不正确的问题。
修复 set password 语句当 host 为空时语法报错的问题。
修复 RawString 收到 null 值时的 NPE 问题。
修复 alter table remove partitioning 在主键和拆分键相同时失败的问题。
修复数据迁移过程中，可能因为非预期主键冲突而降低数据迁移速度的问题。
修复大规模 DN 节点同时 HA 时有概率卡顿并失效的问题。
修复聚合函数 sum 执行时的 decimal 数值处理问题。
修复 ENABLE_BACKGROUND_STATISTIC_COLLECTION 配置不生效的问题。
修复执行 show full columns 时暴露隐式主键导致 sharding jdbc 访问失败的问题。
修复表中含有 JSON 类型列时，执行 REPLACE、UPSERT 命令时报错的问题。
修复 found_rows 函数在存储过程中计算结果不正确的问题。
修复列名大写时，统计信息中 topN 丢失的问题。
修复执行 Alter Table 命令重建复合主键并删除部分主键列时，DDL 任务可能失败且无法继续或回滚的问题。
修复 json_extract 函数错误地返回了字符串 "NULL" 的问题。
修复执行 alter table drop column 可能无法访问该表的问题。
修复 AUTO 模式数据库中，包含 GSI 的 BkaJoin 执行过程中， Lookup Key 推导出错的问题。
修复私有协议在后端连接数不足时的报错信息。
修复私有协议在 CN 高负载场景下小概率死锁的问题。
修复归档表查询低概率卡死的问题。
修复私有协议在传输大包时，可能出现探活任务错误终止连接的问题。
修复 DRDS 模式下的多表 join 生成错误的执行计划的问题。
修复 TTL 表 RENAME 之后，TTL 表失效的问题。
修复 varchar 类型的分区键在写入负数的情况下，隐式类型转换处理出错的问题。
修复子查询（subquery）物化（materialize）执行模式估算子查询行数为 500 时，执行计划构建错误的问题。
修复私有协议探活可能延迟 10 分钟进而造成连接抖动的问题。
修复跨库 analyze table 权限校验失败的问题。
禁止 alter table convert to character set。
针对默认主键拆分表，将其全局二级索引的默认分区数由实例级别修改为表级别。