Replication processing consists of multiple stages. For example, in a topology that runs a Server Manager instance on the source system and on the target system, replication processing occurs in the following stages:
The Extractor captures changes from the source database and writes the changes to intermediate files. Then the Extractor sends a message to the Server Manager on the source system that confirms the extraction stage is complete.
The Server Manager instance on the source system notifies the Server Manager on the target system that the intermediate files are ready to be transferred.
The Server Manager instance on the target system uses HTTP or HTTPS to get the intermediate files and then sends a message to the Server Manager on the source system to confirm that the files were transferred.
The Server Manager instance on the source system removes the intermediate files that were transferred to the target.
For replication jobs with multiple targets, the Server Manager instance waits until the intermediate files have been transferred to all of the targets.
The Applier applies the changes from the intermediate files to the target database. When apply processing is complete, the Applier sends a confirmation message to the Server Manager instance on the target. The Server Manager purges the intermediate files for which all changes were processed and retains only the intermediate files that contain uncommitted transactional changes.