tencent cloud

For TencentDB for MySQL, you can modify the parameters of a source instance in the console. Modifying some crucial parameters in an improper way will lead to exceptions in disaster recovery instances and data inconsistency. This document describes the consequences of modifying the following crucial parameters.

lower_case_table_names

Default value: 0
Description: When creating a database or table, you can set whether storage and query operations are case-sensitive. This parameter can be set to 0 (case-sensitive) or 1 (case-insensitive), and the default value is 0.
Impact: After the parameters of the source instance are modified, the parameters of the disaster recovery instance cannot be modified accordingly, as the source instance is set as case-sensitive, but the disaster recovery instance is not; for example, if two tables named "Test" and "TEst" are created in the source instance, then data sync will fail when the disaster recovery instance uses the corresponding logs, because the table name "TEst" already exists.

auto_increment_increment

Default value: 1
Description: It is used as the increment value of the auto-increment column AUTO_INCREMENT. Its value can range from 1 (default value) to 65,535.
Impact: After the parameters of the source instance are modified, those of the disaster recovery instance cannot be modified accordingly. When binlog_format is set as statement, only statement execution is recorded. In this case, if the increment column value of the source instance is modified but that of the disaster recovery instance is not modified accordingly, the data will be inconsistent between the source and slave instances.

auto_increment_offset

Default value: 1
Description: It is used as the start value (offset) of the auto-increment column AUTO_INCREMENT. Its value can range from 1 (default value) to 65,535.
Impact: After the parameters of the source instance are modified, the parameters of the disaster recovery instance cannot be modified accordingly. When the offset value is modified for the source instance but not for the disaster recovery instance, source-replica data inconsistency will occur.

sql_mode

Default value: NO_ENGINE_SUBSTITUTION
Description: TencentDB for MySQL can operate in different SQL modes, which define the SQL syntax and data check that it should support. The default value of this parameter in v5.6 is NO_ENGINE_SUBSTITUTION, which means that if the used storage engine is disabled or not compiled, an error will be reported; in v5.7, the default value is ONLY_FULL_GROUP_BY,STRICT_TRANS_TABLES,NO_ZERO_IN_DATE,NO_ZERO_DATE.
,ERROR_FOR_DIVISION_BY_ZERO,NO_AUTO_CREATE_USER,NO_ENGINE_SUBSTITUTION.
Here:

• ONLY_FULL_GROUP_BY means that in a GROUP BY operation, the column in SELECT or the HAVING or ORDER BY subquery must be a function column that appears in or relies on GROUP BY.
• STRICT_TRANS_TABLES enables strict mode.
• NO_ZERO_IN_DATE indicates whether the month and day of a date can contain 0 and is subject to the status of the strict mode.
• NO_ZERO_DATE means that dates in the database cannot contain zero date and are subject to the status of the strict mode.
• ERROR_FOR_DIVISION_BY_ZERO means that in strict mode, if data is divided by 0 during the INSERT or UPDATE process, an error rather than a warning will be reported, while in non-strict mode, NULL will be returned.
• NO_AUTO_CREATE_USER prohibits GRANT from creating a user whose password is empty.
• NO_ENGINE_SUBSTITUTION means that if the used storage engine is disabled or not compiled, an error will be reported.

Impact: After the parameters of the source instance are modified, the parameters of the disaster recovery instance cannot be modified accordingly. When the SQL mode is changed for the source instance but not for the disaster recovery instance, for example, if the SQL mode limit of the source instance is smaller than that of the disaster recovery instance, the SQL statements that are successfully executed in the source instance may trigger errors when synced to the disaster recovery instance and thus lead to source-replica data inconsistency.