Question 90 A scheduled Lakeflow Connect ingestion succeeds in extraction but the UC target shows zero new rows. Which first check best distinguishes no source changes from target write failure? Delete the target table and rerun Inspect connector/job run metrics and source change counts alongside target write results Change every catalog permission to ALL PRIVILEGES Inspect only the connector credentials while ignoring source and target row metrics Explanation Correct Answer B:Source extraction metrics and destination commits clarify whether zero new rows is expected or a failure. Incorrect Answers A:Destructive reset is not a safe first diagnostic action. C:Unnecessarily broad grants do not diagnose whether the source changed. D:A successful authentication alone cannot distinguish no source changes from a failure to write eligible records.
Question 91 An ingestion job loads files into bronze, and a business user requests a guarantee that the source collection has no missing files. What evidence should be compared? Compare only the total volume of storage bytes without counting published input records Trust a successful Spark task status as proof every expected source object was processed Expected source inventory or feed manifest against recorded successfully ingested files Use only the number of files in the target folder without source-manifest reconciliation Explanation Correct Answer C:An external expected-file record combined with load tracking is required to validate completeness. Incorrect Answers A:Byte size can vary due to compression and row width; it cannot establish that all expected source records were ingested. B:Compute success does not reconcile expected source files or records against actual committed ingestion counts. D:A target file count does not establish whether the expected source records were all included or whether some inputs were skipped.
Question 92 A cloud folder holds many input file types. COPY INTO must load only JSON documents for one target. Which setting controls the format interpretation? Spark driver maximum cores alone FILEFORMAT = JSON (and an appropriate source filter) An unrelated GRANT MODIFY statement A VACUUM retention interval Explanation Correct Answer B:The file-format option tells COPY INTO how to interpret selected JSON source objects. Incorrect Answers A:Driver resources cannot declare source format semantics. C:MODIFY permits writes but does not specify JSON parsing. D:Vacuum retention manages old Delta files, not JSON input decoding.
Question 93 An ingest process must expose new records within minutes after they arrive, but the current batch runs only once per day. What change directly addresses the freshness gap? Increase dashboard refresh frequency but leave the source ingestion on a daily batch schedule Use a more frequent incremental or streaming ingestion design with corresponding orchestration Change table names from lowercase to uppercase Disable source authentication Explanation Correct Answer B:Freshness improves when source discovery and processing are scheduled or triggered within the SLA. Incorrect Answers A:The report would poll stale output because no new source records enter the target until the next daily ingestion run. C:Identifier casing does not shorten the source-to-table delay. D:Removing authentication is insecure and does not create a reliable data freshness solution.
Question 94 A CDC feed includes deletes as well as inserts. The target represents current entity state. What is incorrect about using append-only INSERT for every event? Append-only writes inherently apply upstream deletes Changing job start time automatically applies CDC deletes Delete events require state-aware handling, such as keyed MERGE or delete processing A source delete should always be interpreted as a new active record Explanation Correct Answer C:Append-only inserts cannot remove source-deleted entities from a current-state projection. Incorrect Answers A:Without explicit logic, inserted events do not delete target rows. B:Scheduling does not enforce event semantics in target data. D:This would invert the source state and corrupt the current-state table.
Question 95 A source sends customer updates and occasional corrected historical rows. The target must apply only the latest version per ID. Which field is most important to retain? A reliable source sequence/version or update timestamp alongside the business key Use the destination Delta commit timestamp as the only customer update ordering field The random order of files returned by listing Record only the timestamp at which the Databricks job happened to load each record Explanation Correct Answer A:A version ordering signal supports choosing the latest state and reconciling out-of-order corrections. Incorrect Answers B:A processing commit timestamp can differ from the source change sequence, especially when older corrected events arrive late. C:File listing order is not a trustworthy source state chronology. D:Load timestamps reflect pipeline arrival, not authoritative source revision order; delayed corrections could overwrite newer business events.
Question 96 A development source sends a batch file twice after a network retry. The destination needs exactly one copy per immutable file in the basic file-load design. What helps? Using a different file path alias on every attempt An idempotent COPY INTO load against the same source files Deleting the target before every retry Using INSERT INTO SELECT over all files on every retry Explanation Correct Answer B:COPY INTO normally tracks loaded files and skips repeat ingestion of the same file. Incorrect Answers A:Changing source identity can undermine stable file tracking rather than ensure deduplication. C:A reset is not the same as incremental exactly-once file ingestion. D:Repeated unguarded selects can reappend file contents.
Question 97 A team has file records that include personally identifiable fields. The ingestion service should preserve records, but only authorized identities may access the resulting table. What is required? Use appropriate UC table and namespace privileges on the landed data Disable source file auditing Protect source file access but leave the resulting table SELECT grant open to all workspace users Give every workspace user SELECT on all catalogs Explanation Correct Answer A:Governed table access prevents the ingestion mechanism from implicitly making sensitive rows public. Incorrect Answers B:Removing audit visibility would weaken accountability rather than restrict reads. C:Restricting raw source paths does not enforce the requested access limits on the separately governed table after ingestion. D:A blanket grant contradicts least-privilege access to sensitive records.
Question 98 A managed source connector periodically takes a snapshot, but an analyst expects 1-minute freshness. What should be confirmed before promising this? That all connectors provide identical low-latency CDC Assume the connector always supports a one-minute schedule because the target SQL query is fast That Delta tables always ingest continuously without a job The connector's supported refresh and source-change capture behavior Explanation Correct Answer D:Managed connector availability does not imply any arbitrary latency SLA; source and connector cadence matter. Incorrect Answers A:Connectors differ by supported source, integration and update semantics. B:Fast query serving does not imply the upstream managed connector supports one-minute capture or source refresh. C:A Delta target does not itself extract records from external sources.
Question 99 An ingestion API produces multiple source records with the same business key but different update times. The silver model requires the newest record. Which operation is relevant? Keep whichever row happens to be returned first by the API Use count(*) to choose which row is newer Deduplicate or merge by business key ordered by a trustworthy source update field Use a new target table for every API page without reconciling duplicate business keys Explanation Correct Answer C:Keyed ordering identifies the latest valid source state before silver publication. Incorrect Answers A:Source response order is not necessarily a reliable event-time ordering. B:Row count cannot select latest event state for an individual entity. D:Separate page tables merely postpone the key conflict and cannot enforce one authoritative row per business key.
Question 100 A team wants to ingest a proprietary source via Lakeflow Connect but the required connector is not supported. What is the practical next assessment? Evaluate an approved partner connector or authenticated custom extraction path Configure a connector for an unrelated supported source and assume it can read the proprietary source Assume any managed connector can connect to every database Assign the extraction principal broader Databricks workspace permissions without implementing source authentication Explanation Correct Answer A:Connector selection must respect actual supported sources; partner or custom integrations can address gaps. Incorrect Answers B:Connector support is specific to each source system; choosing a different vendor connector cannot create compatibility with an unsupported source. C:Managed connectors support defined source systems rather than all arbitrary endpoints. D:Workspace role changes cannot create support for an unavailable source connector or replace a source-specific authenticated extraction method.