100% Latest & Updated Amazon AWS Certified Data Engineer - Associate DEA-C01 Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 370
An Athena Spark notebook reads only a handful of development files yet a manager interprets its average as an entire production trend. Which technical dependency most directly explains this behavior?
- The sample does not represent the target population and interval.
- The data quality contract omits numeric precision and tolerance.
- Credential retrieval and rotation are not part of the workload lifecycle.
- The pipeline has no required-field data quality gate before publishing.
Explanation
Correct Answer
A:The sample does not represent the target population and interval. The corresponding fix is to define a representative sampling plan or full-data check and report uncertainty.
Incorrect Answers
B:This diagnosis addresses another AWS case: An upstream system changes a decimal field from two to four digits of precision and rounding introduces revenue drift. The dependency here is different.
C:This diagnosis addresses another AWS case: A scheduled ETL uses an API token embedded in source code; rotation breaks the batch without an alert. The dependency here is different.
D:This diagnosis addresses another AWS case: A batch containing a required `customer_id` null value is committed to Redshift and breaks joins the next day. The dependency here is different.
Question 371
A Glue streaming job writes duplicate target rows when a microbatch is retried after a worker restart. Select TWO complementary measures that correctly repair and validate this failure.
- Use stable event identities and a sink merge/conditional-write design tied to checkpoint semantics.
- Keep transformations distributed and collect only bounded diagnostic samples.
- Delete streaming checkpoints each time duplicates appear.
- A forced retry leaves one intended row per business event.
- Transient injected failures recover while invalid source data stops deterministically.
Explanation
Correct Answers
A:Use stable event identities and a sink merge/conditional-write design tied to checkpoint semantics. The underlying cause is: The external sink operation is not idempotent across checkpoint replay.
D:A forced retry leaves one intended row per business event. This is the observable result required before accepting the correction.
Incorrect Answers
B:That corrective action is for a different case: A distributed Spark process collects a large dataset to the driver and crashes from memory exhaustion. It does not resolve this incident.
C:Delete streaming checkpoints each time duplicates appear. That shortcut conflicts with the stated requirement.
E:This would validate a different case: A Step Functions state retries every failure including invalid schema, consuming budget without possibility of success. It does not verify the outcome required here.
Question 372
After resolving the following incident, which ongoing safeguard most directly prevents recurrence? A data transformation uses one Spark partition for a terabyte input and only one executor stays busy.
- Version binary-format contracts and include golden sample fixtures.
- Design outputs around replay-safe execution IDs and partial-failure tests.
- Inspect Spark physical plans and skewed partitions after release.
- Apply schema and timestamp quality tests at ingress.
Explanation
Correct Answer
C:Inspect Spark physical plans and skewed partitions after release. This safeguards against a recurring cause: The processing layout prevents distributed parallelism.
Incorrect Answers
A:That safeguard addresses another case: A streaming producer writes compressed messages into a format the Lambda consumer assumes is plain UTF-8. It is not the control for this recurring failure.
B:That safeguard addresses another case: An orchestrator invokes a Lambda that writes to S3 then times out; the retry writes a second output under a new random key. It is not the control for this recurring failure.
D:That safeguard addresses another case: A data ingestion service accepts API JSON but passes inconsistent event time zones to downstream partitions. It is not the control for this recurring failure.
Question 373
An S3 event notification triggers Glue for every file of a large export, producing hundreds of competing jobs. Select THREE complementary measures that correctly repair and validate, and prevent recurrence of this failure.
- Assume S3 notifications arrive exactly once and in global order.
- Aggregate object events or publish a manifest/ready signal and start bounded ETL batches.
- A forced mid-run failure leaves the previous published partition available.
- Define a producer completion contract and deduplicate notifications.
- One controlled export yields the intended bounded job count and all objects are processed.
- Evaluate a broadcast join or other optimized join strategy within Glue/Spark memory limits.
Explanation
Correct Answers
B:Aggregate object events or publish a manifest/ready signal and start bounded ETL batches. The underlying cause is: The event-to-job design starts a run per object without a batch completion boundary.
D:Define a producer completion contract and deduplicate notifications. This protects against the following risk: The event-to-job design starts a run per object without a batch completion boundary.
E:One controlled export yields the intended bounded job count and all objects are processed. This is the observable result required before accepting the correction.
Incorrect Answers
A:Assume S3 notifications arrive exactly once and in global order. That shortcut conflicts with the stated requirement.
C:This would validate a different case: A Glue workflow reruns yesterday's partition after a partially failed daily job but overwrites already-approved historical results. It does not verify the outcome required here.
F:That corrective action is for a different case: A Glue job joins a tiny lookup table to billions of rows and performs a full distributed shuffle. It does not resolve this incident.
Question 374
After resolving the following incident, which ongoing safeguard most directly prevents recurrence? A monitoring rule sees one Athena query failure and pages the full team despite a known transient backend retry.
- Define alarm thresholds using service objectives and error budgets.
- Test pagination and retries against large synthetic bucket prefixes.
- Require multidimensional quality gates before final publish.
- Document allowed lateness and watermark monitoring by source.
Explanation
Correct Answer
A:Define alarm thresholds using service objectives and error budgets. This safeguards against a recurring cause: Alarm logic lacks an agreed signal threshold and actionable severity.
Incorrect Answers
B:That safeguard addresses another case: A daily Python ingestion script calls S3 ListObjectsV2 only once and misses most keys after data growth. It is not the control for this recurring failure.
C:That safeguard addresses another case: A data team declares a successful cleanup because duplicate percentage fell, but the job also dropped valid unique rows. It is not the control for this recurring failure.
D:That safeguard addresses another case: A nightly data product uses a moving seven-day window but incoming source events arrive late and are excluded without a watermark policy. It is not the control for this recurring failure.
Question 375
A Redshift view exposes unmasked customer SSNs to a general BI role despite a requirement for partial disclosure. Which engineering change best addresses the underlying issue?
- Restrict cross-Region replication and backup destinations to approved residency Regions.
- Apply supported dynamic data masking or carefully governed masked views and role access.
- Redact values at source and restrict log retention and access by least privilege.
- Grant scoped Lake Formation SELECT and required catalog permissions to the authorized principal.
Explanation
Correct Answer
B:Apply supported dynamic data masking or carefully governed masked views and role access. It directly addresses the case cause (The published data access layer does not implement required masking or view restrictions.) and should be verified by: An ordinary BI user sees only the approved masked representation.
Incorrect Answers
A:That change addresses a data lake in an EU Region replicates raw customer records to an unapproved Region for convenience; it would not repair this incident.
C:That change addresses a central CloudWatch Logs group for Glue jobs retains all raw customer payloads in debug messages across many teams; it would not repair this incident.
D:That change addresses a Lake Formation-enabled S3 table is visible in Glue Data Catalog, but Athena SELECT is denied to the analyst role; it would not repair this incident.
Question 376
A team applies a reviewed fix for this situation: An EventBridge schedule starts a warehouse load during an upstream S3 export rather than after all files are ready. Which observed outcome would most convincingly validate the fix?
- A clean deployment reproduces expected resources and scoped permissions.
- Duplicate notifications create one committed warehouse batch and auditable duplicate attempts.
- Multiple tasks use the available cluster and total runtime improves.
- The load starts only when the reviewed export completion condition is met.
Explanation
Correct Answer
D:The load starts only when the reviewed export completion condition is met. This verifies the result of use an explicit completed-manifest event or a sensor/check against expected export state.
Incorrect Answers
A:This would verify another situation: A serverless pipeline deploys with hand-edited IAM and S3 settings that differ among test and production accounts. It does not validate the current case.
B:This would verify another situation: A production Glue workflow receives the same completed-file notification twice and creates two warehouse loads. It does not validate the current case.
C:This would verify another situation: A data transformation uses one Spark partition for a terabyte input and only one executor stays busy. It does not validate the current case.
Question 377
A QuickSight dashboard compares two branch cost reports, one in local currency and one in USD, without conversion. Which engineering change best addresses the underlying issue?
- Move bulk transformation to a suitable managed batch/distributed service and orchestrate it.
- Run a null/uniqueness check in Glue or SQL and quarantine failing rows or batches.
- Normalize currencies using approved rates and label dates, units and calculation rules.
- Check expected row counts against source manifest and tolerances before release.
Explanation
Correct Answer
C:Normalize currencies using approved rates and label dates, units and calculation rules. It directly addresses the case cause (Visualization dimensions and units are inconsistent between sources.) and should be verified by: A known fixture yields equivalent converted branch costs and a correctly labeled chart.
Incorrect Answers
A:That change addresses a monthly data load uses Lambda to copy hundreds of gigabytes in a single invocation and repeatedly times out; it would not repair this incident.
B:That change addresses a batch containing a required `customer_id` null value is committed to Redshift and breaks joins the next day; it would not repair this incident.
D:That change addresses a feed arrives with 30% fewer rows than normal, but a transformation succeeds and marks the data product complete; it would not repair this incident.
Question 378
A Glue processing library has no pinned dependency versions and a release pulls a new incompatible parser. Which engineering change best addresses the underlying issue?
- Consume a supported completion event or atomic finalized object naming before ingestion.
- Choose a stable order identity and the correct aggregation or distinct-current-state logic.
- Pin and test dependency versions within the supported Glue runtime and publish a locked artifact.
- Model all documented terminal statuses with success, retryable, and failure branches.
Explanation
Correct Answer
C:Pin and test dependency versions within the supported Glue runtime and publish a locked artifact. It directly addresses the case cause (Uncontrolled dependency drift changes execution behavior.) and should be verified by: A rebuild reproduces the tested parser and expected input transformation.
Incorrect Answers
A:That change addresses a partner uploads a CSV over SFTP to Transfer Family but a load starts before the file finishes writing; it would not repair this incident.
B:That change addresses a nightly aggregation counts an order once per status-change event rather than once per order ID; it would not repair this incident.
D:That change addresses a workflow branches on a job-status string but overlooks the Cancelled terminal state, leaving executions stuck; it would not repair this incident.
Question 379
A team applies a reviewed fix for this situation: An SQS-backed processor overwhelms a downstream database after scaling consumers without any concurrency cap. Which observed outcome would most convincingly validate the fix?
- A fixture with two nested lines produces correct line and order totals.
- Rebalances settle and no duplicate record creates a second downstream business event.
- Backlog declines within SLA without excessive database connection or write failures.
- A replay starting at the intended timestamp retrieves the retained events.
Explanation
Correct Answer
C:Backlog declines within SLA without excessive database connection or write failures. This verifies the result of set supported concurrency/batch controls and scale throughput using measured sink limits.
Incorrect Answers
A:This would verify another situation: A Glue job reading nested JSON flattens arrays incorrectly and duplicates parent order totals. It does not validate the current case.
B:This would verify another situation: An MSK consumer group repeatedly rebalances during long processing, causing duplicate deliveries. It does not validate the current case.
D:This would verify another situation: A fraud detector needs to reread yesterday's Kinesis records but the stream uses its short default retention window. It does not validate the current case.
Question 380
A team applies a reviewed fix for this situation: A Glue job uses a fixed worker count at peak and wastes compute on tiny nightly delta batches. Which observed outcome would most convincingly validate the fix?
- Small batches finish at lower cost without violating latency objectives.
- A maximum-size fixture completes without filling local temporary storage.
- A load rehearsal stays within accepted throttle and freshness targets.
- Observed record-to-S3 latency meets the documented achievable target.
Explanation
Correct Answer
A:Small batches finish at lower cost without violating latency objectives. This verifies the result of measure runtime and costs by job size and configure eligible autoscaling or differentiated job profiles.
Incorrect Answers
B:This would verify another situation: A Lambda ETL process writes temporary files larger than its configured /tmp capacity and receives storage errors. It does not validate the current case.
C:This would verify another situation: A Lambda ETL function exhausts concurrency during a flash sale, causing fresh events to be throttled. It does not validate the current case.
D:This would verify another situation: A Firehose pipeline is expected to deliver individual records instantly, yet it buffers batches before S3 delivery. It does not validate the current case.
AWS Certified Data Engineer - Associate DEA-C01 Practice Tests