Question 10 A Delta table is registered in the catalog, and a user can discover its metadata but lacks authorization to read the rows. What platform separation explains this? Grant access to the SQL warehouse and assume that also grants SELECT on the catalog table Metadata discoverability and data-read privileges are distinct All discovered tables are automatically readable Increase the warehouse size until the unauthorized query succeeds Explanation Correct Answer B:Catalog discovery does not imply SELECT access to governed table data. Incorrect Answers A:Warehouse use and Unity Catalog data privileges are separate; the principal can execute SQL without being able to read that table. C:Catalog visibility and data access are separately controlled, so discovery alone is insufficient. D:More compute resources cannot change the principal’s object privileges; a missing SELECT grant is not a resource bottleneck.
Question 11 A team needs a stable analytic serving layer but must keep sensitive source data in object storage with an open transactional table format. Which combination most closely matches a lakehouse? Only a local notebook file system One manually shared single-machine SQL database Cloud object storage with Delta tables and managed compute CSV-only object storage used as the entire analytical serving layer Explanation Correct Answer C:The lakehouse approach combines scalable object storage, Delta transactional semantics and independent execution engines. Incorrect Answers A:Local notebook files are not the shared durable lakehouse data-management foundation. B:A standalone database does not capture the shared object-storage and multi-engine model in the question. D:Raw CSV folders can store records, but do not provide the open transactional Delta table semantics and multi-engine managed serving in the requirement.
Question 12 A temporary development compute resource is left idle for long periods after engineers finish notebooks. Which configuration primarily limits that unnecessary spend? Enable autoscaling but leave the development cluster running indefinitely Disable all Unity Catalog grants Increase job retry count indefinitely Configure an appropriate automatic termination interval Explanation Correct Answer D:Auto termination shuts down idle interactive compute after the configured inactivity window. Incorrect Answers A:Autoscaling changes worker capacity while a cluster is active; it does not replace idle auto termination for periods of nonuse. B:Removing data access harms the workflow and does not specifically stop idle compute charges. C:Retries affect failed executions, not the cost of an idle development cluster.
Question 13 A SQL dashboard workload sees frequent concurrent readers but no Python transformations. Which scaling decision best fits that load? Increase the number of scheduled report notebooks without changing SQL warehouse capacity Increase only the driver memory of one existing SQL warehouse Switch the table to an ungoverned CSV file Configure a suitable SQL warehouse size and scaling settings Explanation Correct Answer D:SQL warehouse resources and concurrency behavior should match interactive read demand. Incorrect Answers A:Additional report scheduling creates workload but does not increase the warehouse concurrency capacity needed by SQL clients. B:A driver setting alone does not establish representative warehouse query concurrency and scaling behavior under peak load. C:Removing transactional storage and governance does not solve SQL-serving capacity.
Question 14 A data engineer can query a Delta table with a three-part name without using a cloud-storage path. What names are represented? Metastore, schema, and table Workspace, catalog, and table Catalog, table, and storage credential Catalog, schema, and table Explanation Correct Answer D:Unity Catalog uses a catalog.schema.table namespace to identify governed tables. Incorrect Answers A:The Unity Catalog metastore is the top-level governance container, not the first identifier in a SQL three-part table name. B:The workspace is not one of the three name components, and this expression omits the required schema level. C:A storage credential governs path access but is not the middle schema component in a table identifier.
Question 15 An analytics team wants to reuse one trusted source table from SQL and PySpark without exporting multiple independent CSV copies. Which approach best meets that need? Replace the Delta table with notebook output text Store the table only as an environment variable Reference the same governed Delta table from both workloads Email a CSV extract to each analyst Explanation Correct Answer C:SQL and PySpark can access a shared registered Delta table with consistent governance and committed content. Incorrect Answers A:Notebook display output does not serve as the common reliable transactional dataset. B:Environment variables are not a durable store for tabular analytics data. D:Separate exports become stale and undermine a shared source of truth.
Question 16 A team is deciding between job compute and interactive compute for a scheduled single-purpose batch. Which factor is most decisive? Assign the production batch to a persistent interactive development cluster Whether the scheduled batch depends on interactive notebook session state Unattended lifecycle and per-run resource needs Require a manually opened notebook session to launch every production run Explanation Correct Answer C:Automated batch work benefits from compute provisioned for its run and stopped afterward. Incorrect Answers A:An interactive cluster can run Spark code, but persistent shared compute is a poor match for a transient unattended batch workload. B:A notebook can contain the batch logic, but this is less decisive than whether compute is automatically provisioned and terminated for each unattended run. D:Manual session-dependent execution does not meet the specified unattended per-run production lifecycle.
Question 17 A business needs tables with rollback-friendly commit history and correctness under retries. Which storage table format feature contributes most? Keep a history of successful notebook executions but not Delta commit versions A transaction log that records table changes Use cloud-object modification timestamps as the authoritative transaction history Uncontrolled modification of existing Parquet files Explanation Correct Answer B:Delta transaction metadata identifies committed actions and enables consistent table versioning. Incorrect Answers A:Job run history may show when code ran, but does not identify the exact committed table state available for time travel or restore. C:File timestamps cannot define which group of files constituted one atomic Delta table version. D:Changing files without a transaction protocol risks inconsistent readers and untracked updates.
Question 18 A platform operator wants to minimize the operational effort of manually tuning file layout for frequently queried Delta tables. Which Databricks platform capabilities are relevant? Run manual OPTIMIZE maintenance on an arbitrary fixed schedule Liquid clustering and supported predictive optimization Adding arbitrary partition columns to every table Keeping all files at the smallest possible size Explanation Correct Answer B:These features can simplify data layout management and automatic table-maintenance decisions when supported. Incorrect Answers A:Manual optimization may improve some layouts, but creates operating work and can miss changing query patterns compared with supported automated optimization. C:Indiscriminate partitioning can create excessive partitions and does not guarantee a suitable layout. D:Very small files may degrade planning and I/O efficiency rather than simplify optimization.
Question 19 An engineer wants to compare the impact of compute configurations on an ETL workload rather than guess which cluster is cheapest. What decision is defensible? Always use the smallest single node for distributed ETL Benchmark representative runtime and cost under each suitable compute option Compare only the listed cloud instance hourly prices without running the ETL workload Always select the largest available cluster Explanation Correct Answer B:A workload-specific measurement captures tradeoffs between execution time and billed compute. Incorrect Answers A:A single small node may bottleneck a workload needing parallel data processing. C:Published compute prices alone cannot capture the runtime, parallelism, or total DBU and infrastructure charges under the real ETL load. D:Bigger clusters are not universally faster per dollar and may add unnecessary spending.
Question 20 A group needs an interactive workspace for code development but wants production execution separated to avoid interference. Which arrangement is best? Disable all scheduled job runs Run production only from one engineer's private notebook session Use development compute for iteration and job compute for production runs Keep the production jobs on the developers’ shared cluster while applying stricter workspace permissions Explanation Correct Answer C:Separating compute lifecycles prevents notebook experiments from competing directly with scheduled batch execution. Incorrect Answers A:Eliminating schedules does not satisfy the production execution requirement. B:A personal development session creates availability and operational ownership risks for production. D:Workspace permissions may restrict users, but the production batch still competes with active development compute, defeating the requested execution isolation.