Question 370 A production deployment must run only after a specific version is approved. Which recorded identifier supports this control? The Git commit or release tag selected by the deployment workflow The data table's approximate row count Use an unversioned floating branch reference with no recorded resolved commit The current wall-clock second of a random notebook launch Explanation Correct Answer A:A pinned revision connects the deployed code to a reviewed and repeatable source state. Incorrect Answers B:Data count cannot establish which code version CI deployed. C:A floating branch can advance between review and deployment, so it does not pin the exact approved source revision. D:A timestamp alone does not identify the tested code revision.
Question 371 A bundle specifies a notebook task but deployment succeeds while execution later fails because the notebook path is wrong. What should be improved? Validate and test the referenced artifact path in the selected target before release acceptance Increase cluster workers without changing the path Grant all users SELECT on every table Change the job display name Explanation Correct Answer A:Deployment validity alone may not prove every runtime path resolves as intended. Incorrect Answers B:Compute scale does not create the missing notebook resource. C:Data access cannot fix a nonexistent notebook path. D:A display-name change does not repair the task code reference.
Question 372 A CI pipeline currently publishes to prod on every feature-branch push. The organization requires only reviewed merges to reach production. What needs to change? Run production deployment automatically after every green feature-branch build Gate prod deployment to approved protected-branch or release events Disable all source control checks Run the same deployment twice Explanation Correct Answer B:The automation trigger must reflect review policy rather than unreviewed feature pushes. Incorrect Answers A:A successful feature-branch build is not equivalent to a reviewed merge; the deployment trigger must enforce approved promotion. C:Removing checks contradicts the reviewed-merge requirement. D:Repeated deployment does not add source review protection.
Question 373 A team operates three workspaces with distinct external dependencies. It wants bundle promotion without copying secrets into repository files. What should it use? Unreviewed manual edits to every deployed job Secure environment-specific references to credentials plus target variables A single public plaintext token for every workspace Put each environment's secrets into a Git-ignored local plaintext settings file on the CI runner Explanation Correct Answer B:Targets provide configuration differences while protected identities/secrets remain outside committed plaintext. Incorrect Answers A:Manual editing increases drift and does not protect credentials. C:A common exposed secret undermines isolation and creates broad compromise risk. D:An ungoverned local secrets file may leak and cannot replace approved managed identity and secret references.
Question 374 A release pipeline validates a bundle, deploys it, and needs to confirm the correct job definition exists. Which post-deployment evidence is relevant? Inspect the deployed target resource identifiers and configuration Only a notebook name appearing in the local editor Treat bundle validate success alone as proof the intended production resource was deployed Treat the presence of the release tag in Git as confirmation of the live Databricks job Explanation Correct Answer A:Readback of the actual job in the intended workspace verifies the declarative deployment effect. Incorrect Answers B:Local source presence does not prove production deployment. C:Validation checks configuration but does not prove which resource was actually deployed to the target workspace. D:A code push records source changes, not the deployment identity or current job version in the workspace.
Question 375 A project wants independent development and production schedules but identical transformation code. What is the minimal sustainable workflow? Promote source code by changing every production table definition in place Maintain one codebase and target-specific job schedule configuration Deploy directly from an unreviewed personal branch Maintain unrelated copies of every notebook in each environment Explanation Correct Answer B:Separating environment configuration from shared code enables reproducible promotions. Incorrect Answers A:Environment promotion should reuse tested transformation code and change target-specific job configuration, not rewrite production table semantics. C:Unreviewed sources weaken production release controls. D:Manual forks encourage drift and inconsistent bug fixes.
Question 376 A bundle codebase was promoted successfully, but a production job runs with a developer service principal lacking target table rights. What must be diagnosed? Recheck only the CI deploy principal privileges and not the job run-as principal The table's row count Check only which CI user deployed the job and ignore the configured runtime identity The deployed job run-as identity and its Unity Catalog privileges Explanation Correct Answer D:Resource deployment and runtime data authorization are distinct; the production identity must have scoped data access. Incorrect Answers A:Deployment can succeed while a distinct runtime principal lacks Unity Catalog access to required tables. B:Data volume does not establish SELECT or MODIFY privileges. C:The deployment principal's rights do not automatically transfer to a different job run-as principal.
Question 377 A team prepares a hotfix and must ensure the same tested code revision reaches prod after staging. What should its CI metadata pin? Only the execution date The latest moving branch tip with no verification A single immutable commit or artifact version across promotion stages Promote the newest branch tip independently in each environment Explanation Correct Answer C:Pinning the same revision prevents drift between what was tested and what was deployed. Incorrect Answers A:A date is not a unique code revision reference. B:A moving tip may differ between staging tests and production deployment. D:Different branch tips can resolve to different code revisions; staging and production should use the same tested immutable source.
Question 378 A Spark stage has 200 tasks, and one runs 20 minutes while the other 199 finish in seconds. Which Spark UI evidence should be inspected first? The number of dashboard visualizations Per-task shuffle read sizes and partition skew for the slow task Look only at the average duration across the 200 tasks Only the notebook owner email Explanation Correct Answer B:One extreme straggler often indicates a partition with disproportionate data or expensive input. Incorrect Answers A:Dashboard counts do not identify task partition skew. C:An average hides the one outlier; per-task shuffle and partition-size evidence is needed to diagnose a straggler. D:Ownership metadata does not reveal stage-level work imbalance.
Question 379 An executor repeatedly spills intermediate data to disk while aggregating a large dataset. What should the engineer infer? The cluster always has too much memory The data must have no shuffle operations The stage has memory pressure relative to its intermediate working set Assume a faulty SQL SELECT grant is causing disk spill Explanation Correct Answer C:Disk spilling indicates execution data exceeded the available in-memory space for that operation. Incorrect Answers A:Spilling generally signals pressure rather than surplus memory. B:Shuffles and large sorts/aggregations often contribute to spilling. D:Authorization failure prevents access; disk spill during an executing aggregation indicates intermediate data exceeding memory availability.
Question 380 A large join stage shuffles 400 GB of data and spends most of its time exchanging partitions. Which optimization should be investigated? Granting every analyst write permissions Join strategy, filters, and partitioning that can reduce the shuffle Increase the number of notebooks in the workspace without changing the join plan Increase the shuffle partition count without inspecting skew, data reduction or the current join strategy Explanation Correct Answer B:Broadcasting a small side, filtering early, or improving partition distribution may reduce exchange cost. Incorrect Answers A:Authorization scope is unrelated to shuffling efficiency. C:Notebook inventory does not reduce network shuffle data exchanged during the large join stage. D:Adding partitions can change task size but cannot by itself remove unnecessary exchanges; diagnose join distribution and filtering first.