Databricks Data Engineer Associate Practice Test

Answer choices

1. A. It is safe as long as editors lack MODIFY on table rows; tags cannot affect visibility.
2. B. It is harmless because ABAC evaluates catalog names, not tags.
3. C. It affects query performance because tags are used solely for clustering.
4. D. Restrict and audit changes to governed tags that drive ABAC policy scope.

Wrong-answer review

• A. It is safe as long as editors lack MODIFY on table rows; tags cannot affect visibility.: Changing policy-driving metadata can alter access independently of row-modification rights.
• B. It is harmless because ABAC evaluates catalog names, not tags.: Governed tags are a core ABAC input, not decorative metadata.
• C. It affects query performance because tags are used solely for clustering.: Tag-based ABAC controls security policy rather than physical data layout.

Answer choices

1. A. Drop and recreate the Unity Catalog metastore because clustering metadata cannot be changed.
2. B. Increase SQL warehouse concurrency because clustering changes are applied by query threads.
3. C. Run `VACUUM` because VACUUM reclusters current data into the new key order.
4. D. Run or allow `OPTIMIZE` to rewrite existing data under the new clustering layout

Answer choices

1. A. A materialized view chosen specifically because it uses batch rather than streaming semantics.
2. B. A manually refreshed CSV export in workspace files.
3. C. A streaming table maintained by a Lakeflow pipeline.
4. D. A standard view whose query is recomputed from the entire source whenever queried.

Answer choices

1. A. Move to a more customizable standard connector or Structured Streaming pipeline where the required custom logic can run.
2. B. Remain on the managed connector because managed connectors typically allow arbitrary runtime libraries.
3. C. Use a larger SQL warehouse; compute size exposes managed-connector extension hooks.
4. D. Disable Unity Catalog because governance prevents custom connector code.

Answer choices

1. A. Use a Pipeline task that invokes the existing Lakeflow pipeline, then make reconciliation depend on it.
2. B. Copy the pipeline's internal transformation code into a notebook task and maintain both versions.
3. C. Use a SQL alert task as the pipeline executor.
4. D. Use a Dashboard task because dashboards can initiate pipeline updates.

Answer choices

1. A. Remove Git metadata from the bundle so branch state cannot affect deployment.
2. B. Use a production-mode target with the appropriate safeguards and service-principal deployment identity.
3. C. Run all production deployments from a developer's interactive notebook session.
4. D. Use development mode in production because it relaxes validation and therefore prevents CI failures.

Answer choices

1. A. Serverless should be rejected because it cannot scale automatically.
2. B. A SQL warehouse is mandatory for each workload that needs automatic scaling.
3. C. Classic compute is preferable because serverless requires the team to choose worker instance types and autoscaling bounds.
4. D. Serverless compute abstracts provisioning and scaling, so use it when the workload is supported

Answer choices

1. A. Load all columns as strings into silver and rely on Gold dashboards to interpret types at query time.
2. B. Replace each null or failed cast with zero or the current timestamp so no rows are lost.
3. C. Overwrite the bronze table after casting each column with Spark's inferred types so downstream jobs read one copy.
4. D. Normalize empty strings to null, parse and cast the timestamp and amount with explicit rules, quarantine or reject rows missing required keys, and validate the resulting silver schema before promotion.

Answer choices

1. A. Neither table type can be converted to managed under any circumstances.
2. B. Supported external-table conversion does not use MOVE/COPY, while foreign-table conversion requires the documented MOVE or COPY handling.
3. C. Foreign-table conversion is metadata-only, while external Delta conversion typically deletes the source files.
4. D. Both conversions typically require a full manual export to CSV first.

Answer choices

1. A. COPY INTO.
2. B. A managed SaaS connector for the file bucket.
3. C. A continuous Kafka connector.
4. D. Auto Loader is mandatory for any cloud file ingestion.