For an incremental load job, the job stops after a checkpoint is taken. A checkpoint records the point in the change stream where incremental processing left off for recovery purposes.

For a combined initial and incremental load job, initial load subtasks that are running are allowed to run to completion and Initial load subtasks that are not running remain in their current states. For the incremental load portion of the job, a checkpoint is written to the checkpoint file or target recovery table before the job stops. The database ingestion job will not be able to record a checkpoint unless a change record has been processed for at least one of the tables in the job during the first job run after deployment. If a checkpoint is not available, the job resumes processing from the configured restart point, which is the latest available position in the change stream by default.

For an initial load job, any

running

subtasks are allowed to run to completion and then the job stops. Non-running subtasks remain in their current states.