The synchronization point processing methods for the target database include the single-phase commit method and the two-phase commit method.

• Single-phase commit method
  This method executes synchronization point processing with a single commit for the target database.
  
• Two-phase commit method
  This method executes the commit request twice (at two phases) for the target database.
  

You use the commitment_method operand in the import system definition to specify the synchronization point processing method. If you change the synchronization point processing method, you must execute an initial startup of the target Datareplicator.

When a commit for the target database terminates normally, Datareplicator writes the queue file's information on the update information up to the commit point, and then starts the next transaction for processing update information. If an error occurs at the target database, Datareplicator terminates import processing, and then waits for a restart. When the single-phase commit method is used, Datareplicator re-executes the transaction that resulted in the error during the previous commit processing. In this case, Datareplicator might not be able to identify the completion of commit processing at the target database (such as when control is not returned normally to Datareplicator due to a communications error during control return from the target database). When this happens, Datareplicator assumes an error and re-executes the transaction when HiRDB is restarted (rerun). If the mapping key is not unique, duplicate data might be inserted.

The two-phase commit method can prevent such insertion of duplicate data. If your version of the target Datareplicator supports the two-phase commit method, use it (define it for use).