Question 790 A transformation unit test fails in a staging pipeline, but the production bundle is syntactically valid. What should the release gate do? Block production promotion until the failed behavioral test is resolved Delete the failing test and deploy immediately Promote because syntactic validation returned success Change the production target name to hide the failed test Explanation Correct Answer A:Bundle syntax validity cannot override a demonstrated data-transformation regression in tests. Incorrect Answers B:Removing evidence without fixing a real regression breaks the test gate. C:Syntactic validation and functional correctness are different release criteria. D:Renaming does not repair the transformation behavior.
Question 791 A release pipeline reports invalid YAML for a new job task block, and no deployment occurs. What is the narrowest initial correction? Delete the approved Git history Add more workspace grants for the deploy principal without correcting the invalid YAML Fix the YAML structure or indentation and rerun bundle validation Ignore YAML parsing and start the task manually Explanation Correct Answer C:A syntax or schema defect must be corrected before a valid job definition can be deployed. Incorrect Answers A:Repository history is unrelated to this parser error. B:Permission changes cannot parse syntactically invalid configuration. Bundle validation must succeed before deployment is attempted. D:A manual run does not repair the invalid declarative resource definition.
Question 792 A team wants two engineers to deploy their own dev versions of the same bundle without overwriting a shared production job. What naming/deployment property helps? Use the identical production deployment identity for every developer Put test code in a global production notebook path Use isolated development target settings and unique deployment identity paths Use the same development target identity and workspace root path for every engineer Explanation Correct Answer C:Separate target or identity paths let developers test without writing to the production resource state. Incorrect Answers A:Shared remote identity risks one person overwriting another environment's job. B:A shared production path increases collision and accidental promotion risk. D:A shared deployment identity and path let independent development deployments overwrite each other rather than isolate users.
Question 793 A deployed bundle change causes a regression after yesterday's release. A known-good commit was previously tested. What is a controlled rollback? Deploy yesterday's code into an unrelated dev workspace only Edit the live notebook by hand without updating Git Delete all job run history and assume the regression is gone Check out the approved known-good commit and redeploy it to the correct target Explanation Correct Answer D:Versioned redeployment restores the documented configuration and code through the established release path. Incorrect Answers A:A dev deployment cannot by itself roll back the affected production target. B:A manual hotfix leaves production out of sync with the declarative source. C:History removal does not restore earlier code or resource settings.
Question 794 A developer has successfully deployed a bundle defining a job but wants one immediate execution for smoke testing. Which CLI command starts the job resource? `databricks bundle summary -t dev` `git diff --stat` `databricks bundle validate -t dev` `databricks bundle run -t dev <job_key>` Explanation Correct Answer D:After deployment, the run command starts the identified job resource in the selected target. Incorrect Answers A:A summary describes deployed resources without executing them. B:A local source diff does not invoke a Databricks workflow. C:Validation checks configuration and does not start the job.
Question 795 A pipeline can deploy a job but a later smoke test cannot execute it because its CI principal lacks run authorization. What must the workflow provision or assign? Grant only CAN VIEW on the job to the CI identity and assume it can start runs Grant broad table ownership instead of the job-specific CAN MANAGE RUN privilege Check only deployment rights and omit whether the CI principal can trigger the deployed job Job-run permissions for the CI test identity as well as deployment access Explanation Correct Answer D:Deployment authority and permission to execute jobs can be separately controlled. Incorrect Answers A:Viewing configuration and results is not the same as having the right to trigger runs under job permissions. B:Data-object ownership is separate from job run permissions and violates least-privilege CI access. C:Successful deployment authorization does not automatically confer job-run privileges needed by a subsequent smoke test.
Question 796 A reviewer suspects a production bundle resolved `catalog_name` to the development catalog. What evidence should be examined before deployment? Inspect only the repository's default variable values without resolving the production target Only the date of the last Git pull Only the length of the bundled notebook The resolved production-target bundle configuration and supplied variable overrides Explanation Correct Answer D:The effective target configuration reveals which catalog a production resource will actually reference. Incorrect Answers A:Default YAML values can differ from runtime target overrides; the resolved production bundle configuration is necessary evidence. B:The last pull time cannot prove the currently resolved production catalog value. C:Code size provides no reliable evidence of a variable's effective value.
Question 797 A Spark cross join combines a DataFrame with 4 rows and a DataFrame with 3 rows without a join predicate. Assuming no filtering, how many result rows are created? 12 4 3 7 Explanation Correct Answer A:A Cartesian product pairs each of four rows with each of three rows, giving 4×3=12. Incorrect Answers B:Four would preserve only the first input and omit the repeated matches to the other DataFrame. C:Three would preserve only the second input rather than forming all row pairs. D:Adding input counts does not describe the number of Cartesian pairs.
Question 798 A bronze record has a tags array containing ["fraud", "priority"]. Silver must flag rows containing "fraud" without exploding the array into multiple rows. Which DataFrame predicate fits? Use `count(tags)` as a filter on the word Use `drop(tags)` before filtering Use `explode(tags)` and assume every record still has one row Use `array_contains(tags, "fraud")` Explanation Correct Answer D:Array membership testing can filter relevant source rows without changing their granularity. Incorrect Answers A:Count does not test whether the specific array member fraud occurs. B:Removing the field loses the information needed to detect the tag. C:Explode changes the row grain and can create multiple rows per original event.
Question 799 A gold report groups invoice rows into monthly buckets, and data spans December 2025 and December 2026. What grouping key prevents different years from being combined? The order of rows in each source CSV A month-granularity date containing both year and month Only the integer month number 12 Only the weekday of the invoice date Explanation Correct Answer B:A year-month key separates December 2025 from December 2026 while retaining monthly grouping. Incorrect Answers A:File order is not a stable calendar grouping key. C:Month alone combines dates from different years and corrupts year-specific totals. D:Weekdays do not identify a month and overlap across many dates.
Question 800 A reporting DataFrame groups sales by store and category but also needs subtotals by store and an overall total in one aggregation. Which Spark operation provides hierarchical subtotals? Use `rollup(store, category)` with the chosen aggregates Use a Cartesian join between sales and stores Use a normal `groupBy(store, category)` only Use `dropDuplicates(["store", "category"])` Explanation Correct Answer A:Rollup produces detail groups and hierarchical subtotal levels for the grouping columns. Incorrect Answers B:A cross join creates additional combinations and does not compute hierarchical aggregation totals. C:Ordinary grouping provides details but not the requested subtotals and grand total levels. D:Deduplication discards records instead of computing sums and subtotals.