Certified Data Engineer Associate Practice Test 5
Question 40
At a medical device company, a nightly batch transform starts a costly interactive cluster and leaves it running idle after the successful run. The dataset `program_enrollments` is in scope, and the control board requires an auditable data lineage. Before release, which acceptance test would actually demonstrate that the root problem is controlled?
- Check that the intended configuration exists for `program_enrollments` and that a healthy run succeeded, without inducing the earlier failure.
- Schedule two equivalent batches and check task results, runtime startup, and cost until cluster termination
- Run only a normal, unchanged `program_enrollments` sample and review DBU cost per successful batch.
- Review a platform-level availability report for `program_enrollments` without testing whether the scheduled pipeline inherits a persistent development compute lifecycle instead of an on-demand job lifecycle.
Explanation
Correct Answer
B:Schedule two equivalent batches and check task results, runtime startup, and cost until cluster termination. This reproduces the scenario-dependent condition, verifies DBU cost per successful batch, cluster lifetime, job status, and library compatibility, and directly tests whether the scheduled pipeline inherits a persistent development compute lifecycle instead of an on-demand job lifecycle is resolved. Observed results, not intent, determine approval.
Incorrect Answers
A:Check that the intended configuration exists for `program_enrollments` and that a healthy run succeeded, without inducing the earlier failure. Installed configuration and a healthy run are insufficient to prove the failure mode is gone. The original issue was the scheduled pipeline inherits a persistent development compute lifecycle instead of an on-demand job lifecycle.
C:Run only a normal, unchanged `program_enrollments` sample and review DBU cost per successful batch. A normal sample does not challenge the scheduled pipeline inherits a persistent development compute lifecycle instead of an on-demand job lifecycle.
D:Review a platform-level availability report for `program_enrollments` without testing whether the scheduled pipeline inherits a persistent development compute lifecycle instead of an on-demand job lifecycle. Aggregate availability does not establish the behavior of affected `program_enrollments` records. It cannot replace the direct exercise: schedule two equivalent batches and check task results, runtime startup, and cost until cluster termination.
Question 41
At an agricultural sensor program, an archive of weekly invoice CSV files is appended once per week, but the team proposes an always-running stream. The dataset `population_records` is in scope, and the target must retain every valid late-arriving event. What evidence should the engineer collect to establish the cause rather than a coincidental symptom?
- Examine only a normal successful run of `population_records` and trend weekly file manifest without recreating the fault.
- Correlate weekly file manifest, ingestion run watermark across affected and unaffected `R034-CASE-009` runs.
- Compare the saved configuration with the signed change ticket for `population_records` while omitting source and output records.
- Review aggregate cluster availability and job completion for `population_records` without correlating the observed error path.
Explanation
Correct Answer
B:Correlate weekly file manifest, ingestion run watermark across affected and unaffected `R034-CASE-009` runs.. By joining weekly file manifest, ingestion run watermark, inserted row count, and duplicate invoice IDs across the affected and unaffected conditions, the investigation directly tests the recorded mechanism: the source arrives in bounded weekly batches and has no continuous low-latency requirement. It also sets a measurable baseline for run the weekly ingestion twice with no new files and confirm the second run adds zero rows.
Incorrect Answers
A:Examine only a normal successful run of `population_records` and trend weekly file manifest without recreating the fault. A healthy-run trend can be useful, but it does not expose the source arrives in bounded weekly batches and has no continuous low-latency requirement. The failing-path weekly file manifest, ingestion run watermark, inserted row count, and duplicate invoice IDs has to be correlated before attributing cause.
C:Compare the saved configuration with the signed change ticket for `population_records` while omitting source and output records. A desired configuration and approval ticket are not runtime proof. In this case they omit weekly file manifest, ingestion run watermark, inserted row count, and duplicate invoice IDs, the direct means of distinguishing the source arrives in bounded weekly batches and has no continuous low-latency requirement.
D:Review aggregate cluster availability and job completion for `population_records` without correlating the observed error path. A successful aggregate platform state can coexist with the source arrives in bounded weekly batches and has no continuous low-latency requirement. This check never observes the affected `population_records` records or their causal path.
Question 42
At an agricultural sensor program, an archive of weekly invoice CSV files is appended once per week, but the team proposes an always-running stream. The dataset `population_records` is in scope, and the target must retain every valid late-arriving event. Which platform configuration or data-engineering change is most defensible?
- Use repeatable scheduled incremental file ingestion to a governed Delta bronze table and record the processed-file boundary
- Keep a 24-hour streaming cluster polling an unchanged folder; use that approach for `population_records`.
- Re-read and append the entire historical folder every week; use that approach for `population_records`.
- Overwrite the bronze table on every delivery; use that approach for `population_records`.
Explanation
Correct Answer
A:Use repeatable scheduled incremental file ingestion to a governed Delta bronze table and record the processed-file boundary. This directly addresses the source arrives in bounded weekly batches and has no continuous low-latency requirement. The design should be accepted only after run the weekly ingestion twice with no new files and confirm the second run adds zero rows; the related operational boundary is to favor scheduled execution and recorded incremental progress over permanent streaming infrastructure.
Incorrect Answers
B:Keep a 24-hour streaming cluster polling an unchanged folder; use that approach for `population_records`. Continuous polling consumes capacity without increasing the weekly delivery frequency. For `population_records`, this cannot remove the causal limitation: the source arrives in bounded weekly batches and has no continuous low-latency requirement.
C:Re-read and append the entire historical folder every week; use that approach for `population_records`. Unbounded reprocessing creates duplicate invoices unless separately deduplicated. For `population_records`, this cannot remove the causal limitation: the source arrives in bounded weekly batches and has no continuous low-latency requirement.
D:Overwrite the bronze table on every delivery; use that approach for `population_records`. Overwrites discard historical rows that incremental consumers still require. For `population_records`, this cannot remove the causal limitation: the source arrives in bounded weekly batches and has no continuous low-latency requirement.
Question 43
At an agricultural sensor program, an archive of weekly invoice CSV files is appended once per week, but the team proposes an always-running stream. The dataset `population_records` is in scope, and the target must retain every valid late-arriving event. A teammate proposes: "Keep a 24-hour streaming cluster polling an unchanged folder." Which technical limitation is the decisive reason not to approve that specific proposal?
- Overwrites discard historical rows that incremental consumers still require. This would apply to another option for `population_records`.
- Unbounded reprocessing creates duplicate invoices unless separately deduplicated. This would apply to an alternative change for `population_records`.
- Continuous polling consumes capacity without increasing the weekly delivery frequency. The proposal does not repair `population_records`.
- Insufficient executor memory is the primary fault; source-file handling and ingestion checkpoints are correct.
Explanation
Correct Answer
C:Continuous polling consumes capacity without increasing the weekly delivery frequency. The proposal does not repair `population_records`. This identifies the known limitation of the exact proposed action. Continuous polling consumes capacity without increasing the weekly delivery frequency. The underlying problem remains the source arrives in bounded weekly batches and has no continuous low-latency requirement; Use repeatable scheduled incremental file ingestion to a governed Delta bronze table and record the processed-file boundary instead targets that failure.
Incorrect Answers
A:Overwrites discard historical rows that incremental consumers still require. This would apply to another option for `population_records`. This limitation belongs to the different alternative "Overwrite the bronze table on every delivery". The stem calls for the specific failure of "Keep a 24-hour streaming cluster polling an unchanged folder".
B:Unbounded reprocessing creates duplicate invoices unless separately deduplicated. This would apply to an alternative change for `population_records`. This failure applies to the separate proposal "Re-read and append the entire historical folder every week".
D:Insufficient executor memory is the primary fault; source-file handling and ingestion checkpoints are correct. The case provides a concrete failing data or control path; a mere alternate theory does not address the source arrives in bounded weekly batches and has no continuous low-latency requirement.
Question 44
At an agricultural sensor program, an archive of weekly invoice CSV files is appended once per week, but the team proposes an always-running stream. The dataset `population_records` is in scope, and the target must retain every valid late-arriving event. What risk-and-cost compromise should the owner accept instead of the proposed shortcuts?
- Re-read and append the entire historical folder every week; monitor results only after launch of `population_records`.
- Overwrite the bronze table on every delivery; monitor results only after launch of `population_records`.
- Favor scheduled execution and recorded incremental progress over permanent streaming infrastructure
- Keep a 24-hour streaming cluster polling an unchanged folder; monitor results only after launch of `population_records`.
Explanation
Correct Answer
C:Favor scheduled execution and recorded incremental progress over permanent streaming infrastructure. This explicitly bounds the compromise while still addressing the source arrives in bounded weekly batches and has no continuous low-latency requirement. Acceptance relies on run the weekly ingestion twice with no new files and confirm the second run adds zero rows rather than simply declaring the change complete.
Incorrect Answers
A:Re-read and append the entire historical folder every week; monitor results only after launch of `population_records`. Unbounded reprocessing creates duplicate invoices unless separately deduplicated. Monitoring after deployment does not make this approach a safe bounded tradeoff for `population_records`.
B:Overwrite the bronze table on every delivery; monitor results only after launch of `population_records`. Overwrites discard historical rows that incremental consumers still require. Monitoring after deployment does not make this approach a safe bounded tradeoff for `population_records`.
D:Keep a 24-hour streaming cluster polling an unchanged folder; monitor results only after launch of `population_records`. Continuous polling consumes capacity without increasing the weekly delivery frequency. Monitoring after deployment does not make this approach a safe bounded tradeoff for `population_records`.
Question 45
At an agricultural sensor program, an archive of weekly invoice CSV files is appended once per week, but the team proposes an always-running stream. The dataset `population_records` is in scope, and the target must retain every valid late-arriving event. Which direct test could falsify the proposed fix instead of only confirming deployment?
- Run only a normal, unchanged `population_records` sample and review weekly file manifest.
- Run the weekly ingestion twice with no new files and confirm the second run adds zero rows
- Review a platform-level availability report for `population_records` without testing whether the source arrives in bounded weekly batches and has no continuous low-latency requirement.
- Check that the intended configuration exists for `population_records` and that a healthy run succeeded, without inducing the earlier failure.
Explanation
Correct Answer
B:Run the weekly ingestion twice with no new files and confirm the second run adds zero rows. This reproduces the scenario-dependent condition, verifies weekly file manifest, ingestion run watermark, inserted row count, and duplicate invoice IDs, and directly tests whether the source arrives in bounded weekly batches and has no continuous low-latency requirement is resolved. Observed results, not intent, determine approval.
Incorrect Answers
A:Run only a normal, unchanged `population_records` sample and review weekly file manifest. A normal sample does not challenge the source arrives in bounded weekly batches and has no continuous low-latency requirement.
C:Review a platform-level availability report for `population_records` without testing whether the source arrives in bounded weekly batches and has no continuous low-latency requirement. Aggregate availability does not establish the behavior of affected `population_records` records. It cannot replace the direct exercise: run the weekly ingestion twice with no new files and confirm the second run adds zero rows.
D:Check that the intended configuration exists for `population_records` and that a healthy run succeeded, without inducing the earlier failure. Installed configuration and a healthy run are insufficient to prove the failure mode is gone. The original issue was the source arrives in bounded weekly batches and has no continuous low-latency requirement.
Question 46
At a municipal transit operator, a point-of-sale application continuously emits order events that downstream fraud monitors must see within minutes. The dataset `housing_applications` is in scope, and an interrupted delivery must be recoverable without data loss. Which correlated observation is the strongest basis for the change decision?
- Correlate event arrival timestamps, source offsets/files across affected and unaffected `R034-CASE-010` runs.
- Review completed run status and aggregate cluster health without record-level correlation
- Compare the current configuration and the signed change approval, without source-row evidence
- Examine only a normal successful run of `housing_applications` and trend event arrival timestamps without recreating the fault.
Explanation
Correct Answer
A:Correlate event arrival timestamps, source offsets/files across affected and unaffected `R034-CASE-010` runs.. By joining event arrival timestamps, source offsets/files, checkpoint state, and bronze freshness lag across the affected and unaffected conditions, the investigation directly tests the recorded mechanism: an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling. It also sets a measurable baseline for force a restart during ingestion and confirm bounded end-to-end latency with no unaccounted source events.
Incorrect Answers
B:Review completed run status and aggregate cluster health without record-level correlation. A successful aggregate platform state can coexist with an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling. This check never observes the affected `housing_applications` records or their causal path.
C:Compare the current configuration and the signed change approval, without source-row evidence. A desired configuration and approval ticket are not runtime proof. In this case they omit event arrival timestamps, source offsets/files, checkpoint state, and bronze freshness lag, the direct means of distinguishing an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling.
D:Examine only a normal successful run of `housing_applications` and trend event arrival timestamps without recreating the fault. A healthy-run trend can be useful, but it does not expose an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling. The failing-path event arrival timestamps, source offsets/files, checkpoint state, and bronze freshness lag has to be correlated before attributing cause.
Question 47
At a municipal transit operator, a point-of-sale application continuously emits order events that downstream fraud monitors must see within minutes. The dataset `housing_applications` is in scope, and an interrupted delivery must be recoverable without data loss. What implementation should the engineer select for the failing workflow?
- Write directly to an ungoverned notebook cache; use that approach for `housing_applications`.
- Schedule one full source scan after business closes; use that approach for `housing_applications`.
- Use a supported incremental streaming ingestion path with durable checkpointing and a governed Delta landing table
- Disable stream checkpoint persistence after a clean run; use that approach for `housing_applications`.
Explanation
Correct Answer
C:Use a supported incremental streaming ingestion path with durable checkpointing and a governed Delta landing table. This directly addresses an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling. The design should be accepted only after force a restart during ingestion and confirm bounded end-to-end latency with no unaccounted source events; the related operational boundary is to pay for incremental processing while maintaining durable recovery and governance.
Incorrect Answers
A:Write directly to an ungoverned notebook cache; use that approach for `housing_applications`. Ephemeral cache data loses durable progress and access controls. For `housing_applications`, this cannot remove the causal limitation: an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling.
B:Schedule one full source scan after business closes; use that approach for `housing_applications`. An end-of-day scan misses the required near-real-time freshness objective. For `housing_applications`, this cannot remove the causal limitation: an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling.
D:Disable stream checkpoint persistence after a clean run; use that approach for `housing_applications`. Without durable progress, restarts can replay events or lose processing state. For `housing_applications`, this cannot remove the causal limitation: an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling.
Question 48
At a municipal transit operator, a point-of-sale application continuously emits order events that downstream fraud monitors must see within minutes. The dataset `housing_applications` is in scope, and an interrupted delivery must be recoverable without data loss. A teammate proposes: "Disable stream checkpoint persistence after a clean run." Which technical limitation is the decisive reason not to approve that specific proposal?
- Without durable progress, restarts can replay events or lose processing state. The proposal does not repair `housing_applications`.
- Ephemeral cache data loses durable progress and access controls. This would apply to another option for `housing_applications`.
- An end-of-day scan misses the required near-real-time freshness objective. This would apply to an alternative change for `housing_applications`.
- Insufficient executor memory is the primary fault; source-file handling and ingestion checkpoints are correct.
Explanation
Correct Answer
A:Without durable progress, restarts can replay events or lose processing state. The proposal does not repair `housing_applications`. This identifies the known limitation of the exact proposed action. Without durable progress, restarts can replay events or lose processing state. The underlying problem remains an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling; Use a supported incremental streaming ingestion path with durable checkpointing and a governed Delta landing table instead targets that failure.
Incorrect Answers
B:Ephemeral cache data loses durable progress and access controls. This would apply to another option for `housing_applications`. This limitation belongs to the different alternative "Write directly to an ungoverned notebook cache". The stem calls for the specific failure of "Disable stream checkpoint persistence after a clean run".
C:An end-of-day scan misses the required near-real-time freshness objective. This would apply to an alternative change for `housing_applications`. This failure applies to the separate proposal "Schedule one full source scan after business closes".
D:Insufficient executor memory is the primary fault; source-file handling and ingestion checkpoints are correct. The case provides a concrete failing data or control path; a mere alternate theory does not address an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling.
Question 49
At a municipal transit operator, a point-of-sale application continuously emits order events that downstream fraud monitors must see within minutes. The dataset `housing_applications` is in scope, and an interrupted delivery must be recoverable without data loss. Which decision preserves the stated data contract while keeping the change scope proportionate?
- Pay for incremental processing while maintaining durable recovery and governance
- Write directly to an ungoverned notebook cache
- Disable stream checkpoint persistence after a clean run
- Schedule one full source scan after business closes; monitor results only after launch of `housing_applications`.
Explanation
Correct Answer
A:Pay for incremental processing while maintaining durable recovery and governance. This explicitly bounds the compromise while still addressing an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling. Acceptance relies on force a restart during ingestion and confirm bounded end-to-end latency with no unaccounted source events rather than simply declaring the change complete.
Incorrect Answers
B:Write directly to an ungoverned notebook cache. Ephemeral cache data loses durable progress and access controls. Monitoring after deployment does not make this approach a safe bounded tradeoff for `housing_applications`.
C:Disable stream checkpoint persistence after a clean run. Without durable progress, restarts can replay events or lose processing state. Monitoring after deployment does not make this approach a safe bounded tradeoff for `housing_applications`.
D:Schedule one full source scan after business closes; monitor results only after launch of `housing_applications`. An end-of-day scan misses the required near-real-time freshness objective. Monitoring after deployment does not make this approach a safe bounded tradeoff for `housing_applications`.
Question 50
At a municipal transit operator, a point-of-sale application continuously emits order events that downstream fraud monitors must see within minutes. The dataset `housing_applications` is in scope, and an interrupted delivery must be recoverable without data loss. What verification provides evidence of correct behavior under the original adverse condition?
- Check that the intended configuration exists for `housing_applications` and that a healthy run succeeded, without inducing the earlier failure.
- Force a restart during ingestion and confirm bounded end-to-end latency with no unaccounted source events
- Review a platform-level availability report for `housing_applications` without testing whether an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling.
- Run only a normal, unchanged `housing_applications` sample and review event arrival timestamps.
Explanation
Correct Answer
B:Force a restart during ingestion and confirm bounded end-to-end latency with no unaccounted source events. This reproduces the scenario-dependent condition, verifies event arrival timestamps, source offsets/files, checkpoint state, and bronze freshness lag, and directly tests whether an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling is resolved. Observed results, not intent, determine approval.
Incorrect Answers
A:Check that the intended configuration exists for `housing_applications` and that a healthy run succeeded, without inducing the earlier failure. Installed configuration and a healthy run are insufficient to prove the failure mode is gone. The original issue was an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling.
C:Review a platform-level availability report for `housing_applications` without testing whether an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling. Aggregate availability does not establish the behavior of affected `housing_applications` records. It cannot replace the direct exercise: force a restart during ingestion and confirm bounded end-to-end latency with no unaccounted source events.
D:Run only a normal, unchanged `housing_applications` sample and review event arrival timestamps. A normal sample does not challenge an end-of-day whole-directory batch cannot satisfy the event-latency bound or robust replay handling.