A company must populate an S3 data lake from an existing supported relational database and then continue ingesting inserts, updates, and deletes with minimal application interruption. Which DMS task type best matches this requirement?

1. A scheduled schema-only export with no row data.
2. CDC only without an existing target baseline, assuming it reconstructs all historical rows automatically.
3. Full load plus change data capture, with the source and target configured for the required change semantics.
4. Full load only, because it automatically continues capturing every future transaction.