AWS Data Engineer Associate Practice Test

Answer choices

1. A. Keep every object in S3 Standard for seven years.
2. B. Transition the objects to S3 Express One Zone.
3. C. Transition the objects to S3 Glacier Deep Archive after the active-retention period.
4. D. Expire the objects after the active-retention period.

Wrong-answer review

• A. Keep every object in S3 Standard for seven years.: S3 Standard preserves frequent-access performance but costs more for an archival access pattern.
• B. Transition the objects to S3 Express One Zone.: Express One Zone is designed for high-performance access in one Availability Zone, not lowest-cost deep archive.
• D. Expire the objects after the active-retention period.: Expiration deletes the records and violates the seven-year retention requirement.

Answer choices

1. A. Use EventBridge Scheduler when the trigger is time based rather than event based.
2. B. Use Amazon SNS when a producer already publishes notifications and no content-based event-bus routing is needed.
3. C. Use Amazon EventBridge with a rule that matches the required event pattern and target.
4. D. Use Amazon SQS when durable queueing/consumer decoupling is the primary requirement rather than event-pattern routing.

Answer choices

1. A. Enable S3 Bucket Keys for the SSE-KMS encrypted bucket.
2. B. Increase the KMS key rotation frequency, which does not reduce the number of S3-to-KMS requests for object encryption.
3. C. Use SSE-S3 instead, abandoning the stated requirement to remain on SSE-KMS with the customer managed key.
4. D. Increase S3 multipart-upload part size and assume fewer object parts eliminate KMS request cost for the workload.

Answer choices

1. A. Enable `BisectBatchOnFunctionError`, which narrows failing batches but can still retry already-successful records and is not the same record-level failure response.
2. B. Enable partial batch response and return the failed record identifiers with ReportBatchItemFailures.
3. C. Reduce batch size to 1 so failures are isolated by brute force, increasing invocation overhead.
4. D. Send the whole failed batch to an on-failure destination after the first error and skip retry of recoverable records.

Answer choices

1. A. Use Glue crawler schema inference and treat a column's inferred type as proof that all records satisfy uniqueness/completeness rules.
2. B. Use AWS Glue Data Quality with rules for allowed ranges and timestamp conditions.
3. C. Implement the checks as ad-hoc Spark assertions in each job with no reusable ruleset or published quality results.
4. D. Use DataBrew interactive profiling only, even though the checks must run automatically in the production ETL pipeline.

Answer choices

1. A. Store the credential in Parameter Store SecureString and implement a custom rotation Lambda/schedule.
2. B. Enable Secrets Manager automatic managed rotation for the supported database secret.
3. C. Rotate only the KMS key that protects the secret while leaving the database credential value unchanged.
4. D. Use a long-lived IAM access key as the database credential and rotate it through IAM.

Answer choices

1. A. Enable partition projection in Athena even though multiple Glue-integrated engines need faster server-side partition filtering from the catalog.
2. B. Create a partition index that reflects the commonly used partition-key filter order.
3. C. Add more crawler runs so partition metadata is refreshed more often, without indexing the large partition catalog.
4. D. Collapse the table to fewer coarse partitions, accepting much larger data scans to reduce catalog partition count.

Answer choices

1. A. Use console screenshots as the authoritative deployment specification.
2. B. Use an S3 Inventory manifest as the infrastructure definition.
3. C. Define the infrastructure with AWS CloudFormation or AWS CDK and deploy it through version-controlled CI/CD.
4. D. Use a Kinesis consumer application to create resources ad hoc at startup.

Answer choices

1. A. Send application logs only to S3 and query them with Athena, accepting delayed/custom log analytics rather than the requested CloudWatch-native operational path.
2. B. Use CloudTrail Lake as the primary destination for application stdout/stderr, even though CloudTrail is oriented to activity/audit events.
3. C. Create CloudWatch metrics only and omit the underlying application log events needed for ad-hoc text/query diagnostics.
4. D. Send the logs to Amazon CloudWatch Logs and use features such as Logs Insights/metric filters as needed.

Answer choices

1. A. Use one bucket policy with application-specific conditions for all ten applications and no access points.
2. B. Create separate S3 buckets for every application and replicate shared data among them.
3. C. Use S3 Object Lambda access points even though no per-request object transformation is required.
4. D. Use separate S3 Access Points with policies for the respective application prefixes.