‹ All DEA-C01 questions

DEA-C01 Practice Question 21: DMS Full Load CDC S3

AWS Certified Data Engineer – Associate (DEA-C01) · Test 3 · question 1 of 10

Data Ingestion and Transformationmedium

An on-premises Oracle database backs a company's order-entry application. The analytics team needs a copy of the order tables in an Amazon S3 data lake that stays within a few minutes of the source. The DBA will not allow full table scans during business hours, and the application code cannot be modified to publish events.

Which approach meets these requirements?

Select one answer
Show answer and explanation

Answer:

  • B. Create an AWS DMS task with the migration type "Migrate existing data and replicate ongoing changes", using the Oracle database as the source endpoint and Amazon S3 as the target endpoint

AWS DMS offers three migration types, and the one described here is full load plus change data capture: "the task migrates existing data and then updates the target database based on changes to the source database." The initial full load seeds the lake, and from then on DMS reads the source's transaction logs, for Oracle, the online and archive redo logs through LogMiner or the binary reader API, rather than re-querying the tables. That is what satisfies both constraints at once: no repeated full scans against the OLTP database, and no application change, because the change stream already exists in the redo logs. S3 is a supported DMS target endpoint, so the CDC output lands directly in the data lake.

Option A copies the wrong artifact. Oracle data files are an internal, engine-specific on-disk format; they are not a readable analytics format, a crawler cannot classify them, and copying live data files produces an inconsistent image anyway.

Option C is exactly the full-table-scan pattern the DBA rejected. Reading every order table every hour over JDBC puts sustained read load on the production database during business hours, and the lake is still up to an hour stale.

Option D misreads which log matters. The Oracle alert log records instance-level diagnostics (startups, errors, structural changes) not row-level DML. Tailing it yields operational messages, not order changes. Change data lives in the redo logs, which are binary and are read through the engine's own APIs, not by a text-tailing agent.

Reference: Creating tasks for ongoing replication using AWS DMS

Practice Test 3 ›

Study guides for DEA-C01