Question 760 An operator must stop an inadvertently started expensive run without deleting its reusable job definition. Which action addresses the active execution? Revoke SELECT from every unrelated analyst group Delete all historical job definitions Destroy the whole workspace Cancel the active job run Explanation Correct Answer D:Cancellation targets the running execution while keeping the saved job and its task configuration. Incorrect Answers A:Unrelated catalog privileges do not stop the specific active run. B:Removing workflows is unnecessary to stop the current execution. C:Workspace deletion is unrelated and far beyond canceling one run.
Question 761 A task invokes a billing API and might time out after the API already accepted an invoice. Before enabling automatic retries, what should be ensured? Increase job retries for every possible API timeout without reconciling whether the invoice was already accepted A stable idempotency key or receipt lookup for the invoice operation Automatic deletion of the entire billing history A fresh random invoice identity on every retry Explanation Correct Answer B:A retry must not create a second charge when the first request succeeded but its response was lost. Incorrect Answers A:Blind retries can create duplicate invoices after an ambiguous timeout; idempotency identity or receipt lookup is required. C:Deleting prior data creates a larger integrity problem and does not prove the API state. D:A new identity makes duplicate billing more likely rather than preventing it.
Question 762 A downstream workflow must start only when an upstream multi-task job finishes successfully, even though the upstream tables might change midway. Which trigger source best expresses the dependency? A trigger on the earliest bronze table update Start the downstream job on any task-completion event from the upstream job Use a time-based schedule at the usual upstream finish time instead of an explicit completion dependency An upstream job completion or Run Job dependency Explanation Correct Answer D:Completion-level dependency prevents premature downstream work on intermediate table updates. Incorrect Answers A:A bronze write can precede required validation and gold publication. B:An individual task can finish while other required tasks still run or fail; the downstream trigger must observe whole-job completion. C:A schedule can start before an unusually slow upstream run finishes and violates the required completion ordering.
Question 763 A workflow has both a schedule and a separate file-driven entry path; the same daily batch can cause two starts. What design consideration is most important? Remove all source delivery logs Assume triggers will never coincide for the same date Choose one authoritative trigger or enforce deduplication and concurrency for shared batch IDs Make all target keys random per invocation Explanation Correct Answer C:Multiple triggers can deliver duplicate invocations that must be reconciled to one intended business processing unit. Incorrect Answers A:Deleting logs conceals the double-trigger instead of preventing it. B:Both mechanisms can fire for the same data unless their relationship is controlled. D:Random keys prevent recognizing duplicated batch work.
Question 764 A Python wheel task must receive a customer region and a replay flag as named arguments so one wheel can serve scheduled and manual runs. Which parameter style is appropriate? Change the cluster's node count to encode the region Configure keyword arguments for the Python wheel task and read them in the wheel entry point Supply only positional arguments even though the wheel task is configured for named keyword arguments Force the wheel to parse only command-line positional arguments regardless of task configuration Explanation Correct Answer B:Wheel tasks can accept key-value parameters when configured for keyword arguments. Incorrect Answers A:Compute configuration is not a business parameter transport mechanism. C:Positional arrays and named key-value parameters are distinct task-configuration modes and should match the wheel entry point. D:Wheel tasks support keyword-argument configuration and need not rely exclusively on positions.
Question 765 An engineer processes three old business dates in separate manual runs. Which run metadata makes the resulting records traceable to the intended partitions? Record only the actual job start timestamp, omitting the requested business partition Persist a random sample row without the date Persist the processing_date job parameter together with the run ID in an audit table Persist only the job display name Explanation Correct Answer C:Recording both the business date and the execution ID links each backfill result to its intended input. Incorrect Answers A:Execution time cannot identify whether this was a current-day run or a backfill of one of the three historical dates. B:A sample row cannot reliably establish which historical partition the run processed. D:A single name is reused for many dates and executions.
Question 766 A file-triggered job occasionally sees the same source object again after a producer re-delivery. The business wants exactly one logical order per order ID. What must the job enforce? Clear the whole target before ingesting each object Use the file arrival timestamp as the sole order key Assume one event notification means one unique order Logical-key deduplication or idempotent target writes independent of the trigger Explanation Correct Answer D:A trigger initiates processing but does not substitute for source-event identity and idempotent ingestion. Incorrect Answers A:Full replacement for one object risks deleting unrelated valid orders. B:Different delivery times can refer to the same logical order. C:Redeliveries can carry the same order even when events are distinct.
Question 767 A bundle declares dev as its default target and prod as a separate target. An engineer runs `databricks bundle deploy` without -t from the bundle root. Which target is selected? No target can ever be selected without -t Both dev and prod on one invocation Prod, because production always overrides defaults Dev, because it is explicitly marked as the default Explanation Correct Answer D:Bundle commands use the configured default target if no alternative target is specified. Incorrect Answers A:An explicitly configured default target permits commands without a target flag. B:A single deploy command does not implicitly deploy all declared targets. C:Production does not automatically become the target without an explicit selection or default.
Question 768 A CI pipeline must deploy a verified release to prod while the project has dev marked as the default bundle target. Which command expresses the intent? `git status` `databricks bundle deploy` `databricks bundle validate` `databricks bundle deploy -t prod` Explanation Correct Answer D:The explicit target prevents accidentally deploying to dev as the default. Incorrect Answers A:Source-control status provides no remote bundle deployment. B:Without an override this bundle selects its dev default. C:Validation checks a configuration but does not perform a prod deployment.
Question 769 A reusable bundle declares a variable `my_cluster_id`. A release pipeline must substitute a particular cluster without editing tracked YAML. What supported mechanism can it use? Pass the cluster ID as an unrecognized positional argument Edit the original commit after its approval Change the source code literal that sets the cluster ID for each environment Pass `--var="my_cluster_id=<id>"` to a bundle command Explanation Correct Answer D:A CLI variable override supplies the target-specific value for the command invocation. Incorrect Answers A:An unrecognized argument does not assign the named bundle variable. B:Mutating approved source history is unnecessary to pass a deployment setting. C:Editing code for every target defeats the declarative variable-override mechanism supported by bundle commands.
Question 770 A CI runner provides a bundle variable named `schema_name` through the environment instead of a CLI flag. Which environment variable naming convention does the CLI support? `SPARK_SQL_schema_name` `BUNDLE_VAR_schema_name` BUNDLE_SCHEMA_schema_name `JOB_TOKEN_schema_name` Explanation Correct Answer B:The BUNDLE_VAR_ prefix identifies an environment-provided bundle variable override. Incorrect Answers A:That string does not populate the Databricks bundle variable mechanism. C:The CLI consumes environment substitutions through BUNDLE_VAR_ followed by the declared variable name, not an arbitrary bundle-schema prefix. D:This is not the documented bundle-variable prefix.