>>>>> "Scott" == Scott Hess <scott@stripped> writes:
Scott> <elble@stripped> wrote:
>> > Yes, there is a huge problem in doing it in BOTH directions. The
>> > infinite loop. You will simply have to recognize that row inserted
>> > (updated) action is due to replication. That means a simple field in
>> > each row could flag it.
>> solution: set SQL_LOG_UPDATE=0 before commiting updates to the other side.
Scott> Keeping in mind that you should also not allow non-update queries to be
Scott> hitting the database you're updating while you have SQL_LOG_UPDATE turned
Scott> down! Otherwise, you'll lose all those "real" updates.
Scott> [Actually, the solution is somewhat simple. Write lock _all_ tables in the
Scott> update log, turn down SQL_LOG_UPDATE, apply the update log, then unlock
SQL_LOG_UPDATE is per thread. This means that if you torn of logging
for yourself, you will not affect the logging from other threads.