When a Replicat group stops in an unclean manner, not all of the threads will reach a common checkpoint position in the trail. Unclean stops can be caused by issuing STOP REPLICAT with the ! option, issuing the KILL REPLICAT command, or by transient errors related to Replicat, the database, or other local processes. You can restore the threads to the same position in the trail after an unclean stop and then start Replicat again from the correct checkpoint position.

In this procedure, the restore position is the high watermark. This is the most recent checkpoint position performed among all of the threads (the last record processed by the fastest thread before the unclean stop). Before starting Replicat, you can make changes to the parameter file, such as to repartition the workload among different or new threads. The repartitioning takes effect in a seamless manner after you start Replicat, because the threads can start from a synchronized state.