Notes for the InnoDB Memcached Binlog:
• Most memcached operations are mapped to DML statements (analogous to insert, delete, update).
Since there is no actual SQL statement being processed by the MySQL server, all memcached
commands (except for flush_all) use Row-Based Replication (RBR) logging. This is independent
of any server binlog_format  setting.
• The memcached flush_all command is mapped to the TRUNCATE TABLE command. Since
DDL commands can only use statement-based logging, this flush_all command is replicated by
sending a TRUNCATE TABLE statement.
• The concept of transactions has not typically been part of memcached applications. We use
daemon_memcached_r_batch_size  and daemon_memcached_w_batch_size 
to control the read and write transaction batch size for performance considerations. These settings
do not affect replication: each SQL operation on the underlying table is replicated right after
• The default value of daemon_memcached_w_batch_size  is 1, so each memcached
write operation is committed immediately. This default setting incurs a certain amount of
performance overhead, to avoid any inconsistency in the data visible on the master and slave
servers. The replicated records will always be available immediately on the slave server. If you set
daemon_memcached_w_batch_size greater than 1, records inserted or updated through the
memcached interface are not immediately visible on the master server; to view these records on the
master server before they are committed, issue set transaction isolation level read
... zobacz całą notatkę