Note:
This section introduces the MySQL to MySQL Link as an example.
1. When DTS executes a full data migration/synchronization, it will read all the data from the source database once, thus increasing the load on the source database. If your database specification is low, we recommend performing the migration/synchronization during off-peak business hours, or reducing the DTS rate before starting the task.
2. Different source database specifications and DTS task configurations result in different impacts on the source database's performance. For example, with source database specifications of 8 cores and 16 GB, the default DTS task uses 8 thread concurrency (which can be adjusted), and in the absence of network bottlenecks, the impacts of the DTS task on the source database's performance are as follows.
During full export phase of the source database, it occupies about 18%-45% of the source database's CPU, increases query pressure of the source database by about 40-60 MB/s, and occupies about 8 active session connections.
During incremental export phase of the source database, there is virtually no pressure on the source database, with only one connection continuously listening to the source database's binlog logs.
3. By default, lock-free migration/synchronization is employed, which means no global lock (FTWRL) is applied to the source database during the task. Only tables without a primary key are added with table locks, others are not locked.
4. When performing data consistency checks, DTS uses the executing account of the task to write into the system database __tencentdb__ in the source database, in order to record data comparison information during the task. Do not delete this system database.
To ensure that data comparison issues can be traced later, the DTS does not delete the __tencentdb__ in the source database after completing the task.
The system database __tencentdb__ occupies a very small space, roughly one thousandth to one ten-thousandth of the source database storage space (for example, if the source database is 50 GB, then the system database __tencentdb__ would be approximately 5 MB-50 MB). It also uses a single-threaded and waiting connection mechanism, so it almost has no impact on the performance of the source database, nor does it pre-empt resources.