Problem
A calculation defect in the buffer for instant change column redo causes a crash.
Scenario that Triggers the Issue
The calculation of log index size must account for changes in column order and record the size of index information in the redo log, as these columns were not previously included. A crash occurs when an UPDATE operation is executed immediately after ALTER TABLE to modify column order, due to miscalculation of redo log space allocation.
Kernel Version Related to the Issue
TXSQL 8.0 versions prior to 3.1.15.002.
Fixed Kernel Version
TXSQL 8.0 3.1.16.003.
Check Method
You can refer to the following command to scan for tables with potential risks.
select replace(name, '/', '.') as 'table_schema.table_name' from information_schema.innodb_tables where TOTAL_ROW_VERSIONS!=0 or INSTANT_COLS!=0;
Fixing Method
2. After the kernel version is upgraded, rebuild and repair corrupted tables using alter table xx engine = innodb;. To avoid business impact caused by table locking, it is recommended to perform repairs during off-peak hours using tools such as pt-osc.