The query of a transaction is replicated by the following flows.
The mode of a replication response has the normal mode and reliable mode.
The normal mode return a response to front end, when a query is executed on the Cluster DB which received the query.
The reliable mode return a response to front end, after waiting to have executed the query in all cluster DBs.
The query from the replication server to each cluster DBs are sent by the thread.
And, the replication server locks other replication request reception by semaphore until all threads are finished.
Even when some transactions are received by multi-master, the order of the query processed during all cluster DBs will be same.
