Question 50 An object-storage source receives a daily folder of immutable Parquet files. Engineers want a simple SQL load that safely skips files already ingested on rerun. Which command fits? INSERT OVERWRITE of the destination every run GRANT SELECT on the source bucket COPY INTO a governed Delta target VACUUM on the source folder Explanation Correct Answer C:COPY INTO tracks previously loaded source files and is retryable and idempotent by default. Incorrect Answers A:Replacing the entire destination risks discarding existing rows rather than incrementally loading new files. B:A privilege is needed for access but does not perform the file ingestion. D:VACUUM removes expired data files from Delta tables; it does not load Parquet records.
Question 51 A batch uses COPY INTO against a cloud folder, and an operator runs the same command again without changing options or adding new files. What is the expected default? The target table is deleted and recreated The command loads the same files again whenever the query executes on a new SQL warehouse Every existing file is appended again Previously ingested files are skipped Explanation Correct Answer D:COPY INTO records what files it has loaded and normally avoids ingesting the same file again. Incorrect Answers A:The command loads the target rather than dropping and replacing it by default. B:COPY INTO records processed file identities at the target; changing the execution warehouse does not force ordinary reruns to ingest known files again. C:Default COPY INTO behavior is idempotent rather than blindly re-appending known files.
Question 52 A previously loaded CSV file in an object-storage folder was edited in place. COPY INTO is rerun normally and the owner expects the changed contents to be reprocessed. What assumption should be corrected? Increasing shuffle partitions forces old files to reload Changing a CSV file automatically performs a Delta MERGE COPY INTO always hashes and overwrites changes to known files COPY INTO normally skips files already registered as loaded, even after in-place modification Explanation Correct Answer D:File-level ingestion tracking means changing the content of a recorded file does not automatically reload it in ordinary mode. Incorrect Answers A:Spark parallelism configuration does not reset COPY INTO ingestion history. B:A source-file edit does not issue target table merge semantics. C:Default idempotent loading is not a general change-detection or upsert solution.
Question 53 A data engineer wants to load selected CSV files matching a naming convention from object storage using SQL. Which COPY INTO capability is directly relevant? DROP TABLE on each unmatched file A source file filter such as a PATTERN or FILES clause REVOKE on the target schema ANALYZE TABLE column statistics Explanation Correct Answer B:COPY INTO supports selecting candidate files using supported file-name filters. Incorrect Answers A:Dropping Delta tables is not a source-file filtering method. C:Revoking rights can block ingestion rather than filter the source filenames. D:Statistics help query optimization but do not select which source files to ingest.
Question 54 Files arrive continuously under a protected ADLS path. A team wants to preserve progress across scheduled restarts of an Auto Loader stream. What is essential? Set the notebook run name to the current date Persist a stable Structured Streaming checkpoint for that ingestion query Keep only a text list of the most recently processed filenames in a notebook variable Recreate a fresh checkpoint directory each run Explanation Correct Answer B:The checkpoint records streaming progress and supports resumable ingestion under supported restart conditions. Incorrect Answers A:Run names do not store streaming offsets or discovered-file progress. C:A notebook variable is transient across restarts and cannot replace a durable Structured Streaming checkpoint. D:A new checkpoint loses the previous progress state and can cause reprocessing.
Question 55 An Auto Loader JSON stream sees an added field under default addNewColumns schema evolution and stops with UnknownFieldException. What should the team do after validating the new field? Switch to a different catalog name to resolve the exception Delete every destination record before restarting Restart the stream using the updated tracked schema and existing checkpoint Assume the stream will never update its schema metadata Explanation Correct Answer C:In addNewColumns mode Auto Loader updates schema metadata then stops; a restart resumes with the evolved schema. Incorrect Answers A:Renaming the catalog is unrelated to the schema-evolution restart process. B:A schema update does not inherently require dropping previously ingested data. D:Default addNewColumns records the additional field before the stop.
Question 56 An Auto Loader stream receives new JSON fields that must be retained for inspection but must not expand the destination schema automatically. Which schema evolution mode fits? Configure schema evolution to add new columns as soon as they arrive Configure failOnNewColumns and permanently halt the ingestion instead of retaining novel fields Rescue unexpected fields into a rescued data column addNewColumns with automatic schema expansion Explanation Correct Answer C:Rescue mode keeps unexpected fields in a structured rescued-data representation without evolving the defined schema. Incorrect Answers A:Automatic column addition changes the defined table schema, contrary to retaining unexpected fields for inspection without expanding it. B:Failing on new columns can enforce a manual schema review but does not meet the request to keep ingesting and preserve unknown values in rescued data. D:That mode adds newly observed fields to tracked schema metadata instead of freezing the schema.
Question 57 A file ingestion must fail when an unexpected new source column appears until a reviewer changes the declared schema. Which Auto Loader mode is appropriate? addNewColumns ignoreCorruptFiles as the only control rescue failOnNewColumns Explanation Correct Answer D:This mode stops for new fields and does not automatically update tracked schema as addNewColumns does. Incorrect Answers A:That option automatically extends schema metadata, contrary to the manual review requirement. B:Ignoring corrupt files does not implement an approval gate for legitimate new columns. C:Rescue records unexpected fields without forcing explicit schema review before continuing.
Question 58 A team receives JSON files containing a nested address object and wants to keep all original attributes in bronze for later flattening. Which ingestion design best preserves evidence? Land raw nested payload with available schema and rescue metadata before silver transformation Flatten all fields by blindly discarding unknown nested keys at the source Flatten all nested address fields before preserving any raw input copy Store only selected leaf attributes while discarding the nested source payload Explanation Correct Answer A:Bronze retains source fidelity so nested fields can be interpreted and validated later. Incorrect Answers B:Dropping unknown fields before bronze can permanently lose source evidence. C:Flattening before bronze preservation can lose unrecognized attributes and obscure the original nested structure on later reprocessing. D:A projected subset may suffice for a narrow report but cannot preserve all original nested source attributes for later modeling.
Question 59 A pipeline must ingest an operational SaaS system using a supported native integration that handles source-specific extraction, configuration and routine updates. Which option should be evaluated? Write a general-purpose REST extractor without confirming the supported connector’s managed change ingestion A suitable Lakeflow Connect managed connector Read a static file export once and treat it as a managed live SaaS connector Implement a generic REST polling script without checking the native connector capabilities Explanation Correct Answer B:Managed connectors offer source-aware ingestion integration for supported enterprise systems. Incorrect Answers A:A custom polling client could retrieve records, but the requirement prioritizes a supported native integration with source-specific maintenance. C:A one-time file read does not implement recurring source-system extraction, change capture, or operational connector management. D:Custom API polling might extract rows, but it lacks the supported connector's source-specific change handling and managed maintenance the requirement prioritizes.
Question 60 A source database publishes frequent changes that must reach Databricks incrementally without full-table copies every hour. What ingestion principle best fits? Reimport the entire database and append it unconditionally every hour Re-export the entire database table on every hourly schedule Use a supported change-aware incremental ingestion/CDC mechanism Use a recurring full-table export but de-duplicate it after loading every complete snapshot Explanation Correct Answer C:Change-aware loads capture source inserts/updates/deletes and avoid unnecessary full snapshots. Incorrect Answers A:Blind full snapshots introduce repeated rows and unnecessary work. B:A repeated full load can work in some settings but does not meet the stated efficient change-aware incremental ingestion goal. D:A reconciliation process can address duplicates but still repeats the full source copy and does not capture only changed rows as requested.