We have a replication environment using dbremote, recently we have notice that some transactions are not received by the remote database. We were able to identify and confirm that transactions are in fact sent to the remote database but the issue is that the prior transaction has a referential integrity error and a rollback is issued, this sound normal however the real issue is that dbremote groups operations to a default of 20 operations to with the intention of speeding up the process by executing less commits. In this case there was a legitimate database error that should not affect that particular remote database but since multiple db operations are grouped in one commit then the roll back is rolling back the db operations for different logical unit of works making the other legitimate transaction that do not have a db error.
-- Why the roll back is affecting other logical units of work? roll back should only rollback the transaction that had the problem.
--Is there any way to force dbremote to do the commits the same way that it was originally done when the database transaction was originally posted, i prefer dbremote to take a little bit longer than missing more transactions that it should
Any help in this regard it would be greatly appreciated