Microsoft DP-700 Practice Test 32
Question 310
A Dataflow Gen2 refresh is canceled after it has written some rows to its destination. What should downstream logic assume?
- Reconcile the destination's partial writes rather than presuming the entire refresh was rolled back.
- Remove the broken destination definition, re-add the data destination, and publish again.
- Reassign or remediate the affected dataflow and connection/staging ownership according to supported guidance.
- Open the dataflow editor validation details rather than only examining previous refresh results.
Explanation
Correct Answer
A:Cancellation can leave writes already made to data destinations, so idempotent recovery is needed. Compare destination row counts with last complete run.
Incorrect Answers
B:The internal destination query is missing, so changing the source schedule will not restore destination metadata.
C:Refresh may depend on credentials or staging artifacts tied to a former user, not only the transformation M logic.
D:Save or validation failures differ from refresh execution failures and require authoring diagnostics.
Question 311
A dataflow displays a red status icon because the last publish validation failed. Which place reveals the issue?
- Remove the broken destination definition, re-add the data destination, and publish again.
- Open the dataflow editor validation details rather than only examining previous refresh results.
- Reconcile the destination's partial writes rather than presuming the entire refresh was rolled back.
- Change the destination query to return a supported table, binary, or null value.
Explanation
Correct Answer
B:Save or validation failures differ from refresh execution failures and require authoring diagnostics. Fix the invalid step and publish successfully.
Incorrect Answers
A:The internal destination query is missing, so changing the source schedule will not restore destination metadata.
C:Cancellation can leave writes already made to data destinations, so idempotent recovery is needed.
D:Destination outputs have supported type requirements; arbitrary lists are invalid for that destination contract.
Question 312
A dataflow output to a warehouse suddenly fails when its destination schema changes. What should be reviewed?
- Change the destination query to return a supported table, binary, or null value.
- Open the dataflow editor validation details rather than only examining previous refresh results.
- Compare current destination schema and column mapping, then correct the configured destination.
- Reconcile the destination's partial writes rather than presuming the entire refresh was rolled back.
Explanation
Correct Answer
C:Destination validation checks compatibility and access at refresh; source data transformation alone may not match new target columns. Test a small representative refresh after remapping.
Incorrect Answers
A:Destination outputs have supported type requirements; arbitrary lists are invalid for that destination contract.
B:Save or validation failures differ from refresh execution failures and require authoring diagnostics.
D:Cancellation can leave writes already made to data destinations, so idempotent recovery is needed.
Question 313
A Spark notebook fails with a driver out-of-memory error after collecting millions of records. What should change?
- Check the attached published Fabric environment's library manifest and active Spark session.
- Use a compatible checkpoint migration or a new checkpoint after validating replay and duplication risks.
- Normalize source schema or apply an explicit supported cast before the Delta read/write contract.
- Replace driver-side collection with distributed DataFrame operations.
Explanation
Correct Answer
D:Collect materializes rows in one driver process; keeping transformations distributed addresses the memory concentration. Rerun with representative large input and monitor driver memory.
Incorrect Answers
A:A package may exist only in staging or another environment; runtime imports depend on the session's published dependencies.
B:Stateful checkpoint formats can be incompatible with structural query changes; blindly reusing the old location may fail.
C:Physical schema inconsistencies can break inference and Delta compatibility; more executors cannot reconcile incompatible types.
Question 314
A notebook errors with AnalysisException: unresolved column after an upstream rename. Which repair is appropriate?
- Update the DataFrame column reference or input schema mapping to match the actual source.
- Inspect skewed partition sizes and adjust partitioning or skew-handling strategy.
- Review effective item and OneLake write permissions on the destination path.
- Use a compatible checkpoint migration or a new checkpoint after validating replay and duplication risks.
Explanation
Correct Answer
A:Schema resolution errors occur before useful processing when referenced columns do not exist in the analyzed plan. Print the DataFrame schema and validate affected expressions.
Incorrect Answers
B:A few oversized tasks can exhaust executors despite sufficient average cluster memory.
C:Storage authorization is separate from DataFrame syntax; a valid computation cannot commit without the necessary write grant.
D:Stateful checkpoint formats can be incompatible with structural query changes; blindly reusing the old location may fail.
Question 315
A notebook cannot import a custom Python library after an environment deployment. Where should the engineer look?
- Normalize source schema or apply an explicit supported cast before the Delta read/write contract.
- Check the attached published Fabric environment's library manifest and active Spark session.
- Inspect skewed partition sizes and adjust partitioning or skew-handling strategy.
- Replace driver-side collection with distributed DataFrame operations.
Explanation
Correct Answer
B:A package may exist only in staging or another environment; runtime imports depend on the session's published dependencies. Start a fresh session and inspect the package version.
Incorrect Answers
A:Physical schema inconsistencies can break inference and Delta compatibility; more executors cannot reconcile incompatible types.
C:A few oversized tasks can exhaust executors despite sufficient average cluster memory.
D:Collect materializes rows in one driver process; keeping transformations distributed addresses the memory concentration.
Question 316
A Spark write fails with a permission denied error on a OneLake path. What should be verified?
- Use a compatible checkpoint migration or a new checkpoint after validating replay and duplication risks.
- Review effective item and OneLake write permissions on the destination path.
- Update the DataFrame column reference or input schema mapping to match the actual source.
- Check the attached published Fabric environment's library manifest and active Spark session.
Explanation
Correct Answer
B:Storage authorization is separate from DataFrame syntax; a valid computation cannot commit without the necessary write grant. Test a minimal write using the notebook's effective identity.
Incorrect Answers
A:Stateful checkpoint formats can be incompatible with structural query changes; blindly reusing the old location may fail.
C:Schema resolution errors occur before useful processing when referenced columns do not exist in the analyzed plan.
D:A package may exist only in staging or another environment; runtime imports depend on the session's published dependencies.
Question 317
An executor repeatedly crashes during a skewed aggregation. Which evidence-guided repair is suitable?
- Update the DataFrame column reference or input schema mapping to match the actual source.
- Move the supported setting to its designated environment or session-start configuration boundary.
- Inspect skewed partition sizes and adjust partitioning or skew-handling strategy.
- Review effective item and OneLake write permissions on the destination path.
Explanation
Correct Answer
C:A few oversized tasks can exhaust executors despite sufficient average cluster memory. Compare worst partition size and executor failures after adjustment.
Incorrect Answers
A:Schema resolution errors occur before useful processing when referenced columns do not exist in the analyzed plan.
B:Some Spark settings cannot be changed after a session is initialized; the correct scope must be chosen.
D:Storage authorization is separate from DataFrame syntax; a valid computation cannot commit without the necessary write grant.
Question 318
A notebook starts successfully but a new Spark configuration key is rejected as read-only. How should it be set?
- Use a compatible checkpoint migration or a new checkpoint after validating replay and duplication risks.
- Move the supported setting to its designated environment or session-start configuration boundary.
- Inspect skewed partition sizes and adjust partitioning or skew-handling strategy.
- Replace driver-side collection with distributed DataFrame operations.
Explanation
Correct Answer
B:Some Spark settings cannot be changed after a session is initialized; the correct scope must be chosen. Restart with the supported configuration and inspect its effective value.
Incorrect Answers
A:Stateful checkpoint formats can be incompatible with structural query changes; blindly reusing the old location may fail.
C:A few oversized tasks can exhaust executors despite sufficient average cluster memory.
D:Collect materializes rows in one driver process; keeping transformations distributed addresses the memory concentration.
Question 319
A Structured Streaming query fails to recover after its grouping state schema changed. What should the engineer consider?
- Review effective item and OneLake write permissions on the destination path.
- Use a compatible checkpoint migration or a new checkpoint after validating replay and duplication risks.
- Replace driver-side collection with distributed DataFrame operations.
- Inspect skewed partition sizes and adjust partitioning or skew-handling strategy.
Explanation
Correct Answer
B:Stateful checkpoint formats can be incompatible with structural query changes; blindly reusing the old location may fail. Test recovery with production-like event replay.
Incorrect Answers
A:Storage authorization is separate from DataFrame syntax; a valid computation cannot commit without the necessary write grant.
C:Collect materializes rows in one driver process; keeping transformations distributed addresses the memory concentration.
D:A few oversized tasks can exhaust executors despite sufficient average cluster memory.
Question 320
A Delta notebook fails reading mixed incompatible Parquet types from source files. What should be repaired?
- Normalize source schema or apply an explicit supported cast before the Delta read/write contract.
- Use a compatible checkpoint migration or a new checkpoint after validating replay and duplication risks.
- Replace driver-side collection with distributed DataFrame operations.
- Inspect skewed partition sizes and adjust partitioning or skew-handling strategy.
Explanation
Correct Answer
A:Physical schema inconsistencies can break inference and Delta compatibility; more executors cannot reconcile incompatible types. Inspect representative source schemas and rerun ingestion.
Incorrect Answers
B:Stateful checkpoint formats can be incompatible with structural query changes; blindly reusing the old location may fail.
C:Collect materializes rows in one driver process; keeping transformations distributed addresses the memory concentration.
D:A few oversized tasks can exhaust executors despite sufficient average cluster memory.
