MySQL uses statement-based logging (SBL), row-based logging (RBL) or mixed-format logging. The type of binary log used impacts the size and efficiency of logging. Therefore the choice between row-based replication (RBR) or statement-based replication (SBR) depends on your application and environment. This section describes known issues when using a row-based format log, and describes some best practices using it in replication.
For additional information, see Section 5.1, “Replication Formats”, and Section 5.1.1, “Advantages and Disadvantages of Statement-Based and Row-Based Replication”.
For information about issues specific to NDB Cluster Replication (which depends on row-based replication), see Known Issues in NDB Cluster Replication.
Row-based logging of temporary tables. As noted in Section 4.1.29, “Replication and Temporary Tables”, temporary tables are not replicated when using row-based format. When using mixed format logging, “safe” statements involving temporary tables are logged using statement-based format. For more information, see Section 5.1.1, “Advantages and Disadvantages of Statement-Based and Row-Based Replication”.
Temporary tables are not replicated when using row-based format because there is no need. In addition, because temporary tables can be read only from the thread which created them, there is seldom if ever any benefit obtained from replicating them, even when using statement-based format.
You can switch from statement-based to row-based binary logging format at runtime even when temporary tables have been created. From MySQL 5.7.25, the MySQL server tracks the logging mode that was in effect when each temporary table was created. When a given client session ends, the server logs a
DROP TEMPORARY TABLE IF EXISTSstatement for each temporary table that still exists and was created when statement-based binary logging was in use. If row-based or mixed format binary logging was in use when the table was created, the
DROP TEMPORARY TABLE IF EXISTSstatement is not logged. In previous releases, the
DROP TEMPORARY TABLE IF EXISTSstatement was logged regardless of the logging mode that was in effect.
Nontransactional DML statements involving temporary tables are allowed when using
binlog_format=ROW, as long as any nontransactional tables affected by the statements are temporary tables (Bug #14272672).
RBL and synchronization of nontransactional tables. When many rows are affected, the set of changes is split into several events; when the statement commits, all of these events are written to the binary log. When executing on the replica, a table lock is taken on all tables involved, and then the rows are applied in batch mode. Depending on the engine used for the replica's copy of the table, this may or may not be effective.
Latency and binary log size. RBL writes changes for each row to the binary log and so its size can increase quite rapidly. This can significantly increase the time required to make changes on the replica that match those on the source. You should be aware of the potential for this delay in your applications.
Reading the binary log. mysqlbinlog displays row-based events in the binary log using the
BINLOGstatement (see BINLOG Statement). This statement displays an event as a base 64-encoded string, the meaning of which is not evident. When invoked with the
--verboseoptions, mysqlbinlog formats the contents of the binary log to be human readable. When binary log events were written in row-based format and you want to read or recover from a replication or database failure you can use this command to read contents of the binary log. For more information, see mysqlbinlog Row Event Display.
Binary log execution errors and replica execution mode. Using
slave_exec_mode=IDEMPOTENTis generally only useful with MySQL NDB Cluster replication, for which
IDEMPOTENTis the default value. (See NDB Cluster Replication: Bidirectional and Circular Replication). When
IDEMPOTENT, a failure to apply changes from RBL because the original row cannot be found does not trigger an error or cause replication to fail. This means that it is possible that updates are not applied on the replica, so that the source and replica are no longer synchronized. Latency issues and use of nontransactional tables with RBR when
IDEMPOTENTcan cause the source and replica to diverge even further. For more information about
slave_exec_mode, see Server System Variables.
For other scenarios, setting
STRICTis normally sufficient; this is the default value for storage engines other than
Filtering based on server ID not supported. You can filter based on server ID by using the
IGNORE_SERVER_IDSoption for the
CHANGE MASTER TOstatement. This option works with statement-based and row-based logging formats. Another method to filter out changes on some replicas is to use a
WHEREclause that includes the relation
@@server_id <>clause with
DELETEstatements. For example,
WHERE @@server_id <> 1. However, this does not work correctly with row-based logging. To use the
server_idsystem variable for statement filtering, use statement-based logging.
RBL, nontransactional tables, and stopped replicas. When using row-based logging, if the replica server is stopped while a replication thread is updating a nontransactional table, the replica database can reach an inconsistent state. For this reason, it is recommended that you use a transactional storage engine such as
InnoDBfor all tables replicated using the row-based format. Use of
STOP SLAVE SQL_THREADprior to shutting down the replica server helps prevent issues from occurring, and is always recommended regardless of the logging format or storage engine you use.