Question 620 A source database has an update timestamp but no guarantee that every row change updates that timestamp. Is a watermark-only incremental extract safe? Yes; target Delta ACID fills in changes the source never emitted Yes; any timestamp column provides a complete CDC feed Yes; increasing the hourly job frequency eliminates missing updates No; the source change indicator must be complete or another capture mechanism is required Explanation Correct Answer D:A watermark that misses some updates cannot guarantee the target sees all changed records. Incorrect Answers A:Target transactions cannot invent missing upstream events. B:An unreliable timestamp does not necessarily change on each source mutation. C:More runs cannot detect source changes not reflected in the extraction predicate.
Question 621 A partner provides an object-storage manifest listing 600 expected files, but only 599 are visible at the deadline. Which release decision best protects completeness? Publish because 599 of 600 files is close enough Hold acceptance and reconcile the missing manifest entry before publishing the batch Drop the manifest from evidence Fabricate the missing file's rows from yesterday Explanation Correct Answer B:The declared manifest and observed files disagree, so the batch is incomplete under the source contract. Incorrect Answers A:An arbitrary completeness tolerance does not satisfy the stated manifest contract. C:Ignoring source delivery evidence conceals the missing input. D:Invented data is not a valid replacement for source delivery.
Question 622 A source publisher delivers a monthly snapshot as a set of parallel files and writes a `_SUCCESS` marker when all files have closed. When should the ingest start? Begin processing when the first partition file appears, ignoring the completion marker Immediately after deleting the marker After verifying the completion marker and its expected file set After the first file appears regardless of other files Explanation Correct Answer C:A source completion marker can define the publish boundary for an immutable multi-file snapshot. Incorrect Answers A:Early ingestion can observe an incomplete monthly snapshot before all parallel writers have finished and emitted the marker. B:Deleting readiness evidence does not guarantee all files are complete. D:Early processing may load only part of the monthly snapshot.
Question 623 An external streaming source replays one event after a producer retry, but its payload includes the same producer event UUID. What should downstream ingest do? Use the stable producer event UUID under a controlled deduplication policy Remove all events from that producer for the entire day Keep both as independent logical events solely because arrival times differ Use a random target UUID and ignore source identity Explanation Correct Answer A:A producer-provided immutable ID distinguishes retry from a separate legitimate event. Incorrect Answers B:Discarding unrelated events is not a narrowly scoped duplicate remedy. C:Retries can have distinct arrival times without representing new business activity. D:New target UUIDs can make identical source events appear different.
Question 624 A feed changes from newline-delimited JSON to a JSON array document. Why might a line-based parser fail? The top-level record framing changed and the reader must match the actual JSON document structure JSON arrays cannot contain structured objects The target Delta transaction log must be deleted Only cloud credentials control JSON parsing Explanation Correct Answer A:JSON lines and a JSON array have different record boundaries that require appropriate parsing options or preprocessing. Incorrect Answers B:Arrays can contain valid objects; the issue is reader framing. C:Log deletion is unrelated and dangerous to correcting JSON framing. D:Credentials govern source access but not document structure interpretation.
Question 625 A daily CSV file includes one grand-total trailer record after all data rows, and an ingest parser treats the trailer as an invoice. What should change? Recognize and validate the trailer separately from business records Delete the invoice ID column Convert the entire source to one text field permanently Apply a normal CSV row parser to the trailer and attempt to cast its total into invoice fields Explanation Correct Answer A:Control totals are file metadata and should be reconciled rather than inserted as invoice rows. Incorrect Answers B:Removing the key cannot prevent misclassifying the trailer record. C:Loss of structured invoice rows is unnecessary to handle a trailer. D:Treating a summary trailer as ordinary business data produces a bogus invoice or a parsing error instead of validated trailer metadata.
Question 626 A daily file provides a trailer with record count and amount sum. What validation adds substantive source assurance? Check only the final table's display order Check only that the filename exists Compare parsed record count and amount sum with the declared trailer totals Check only that Spark launched a task Explanation Correct Answer C:Control-total reconciliation can detect missing, extra, or misparsed records beyond a success flag. Incorrect Answers A:Ordering does not validate source control totals. B:Existence does not prove record completeness or numeric totals. D:Task launch does not prove input-to-output reconciliation.
Question 627 A producer returns HTTP 200 with a JSON body containing an application-level `errors` field and zero accepted records. How should ingestion interpret this? Treat every 200 response as complete accepted data Validate application-level success and accepted-record counts, not HTTP status alone Delete target rows to compensate Append a placeholder accepted record Explanation Correct Answer B:Transport success can coexist with source-business rejection that must be handled explicitly. Incorrect Answers A:HTTP status does not prove business-level processing success. C:Data deletion cannot resolve the source rejection and risks loss. D:Inventing success records conceals the application error.
Question 628 An ingest pipeline uses temporary signed URLs to read cloud objects, and some tasks fail after URLs expire mid-run. What should be fixed? Use a supported credential or URL renewal path with sufficient task validity Rename table columns Increase the number of invalid URLs in the queue Ignore failures and mark the batch complete Explanation Correct Answer A:Authentication material must remain valid during the actual read operations, including delayed retries. Incorrect Answers B:Column naming is unrelated to expiring read credentials. C:More expired URLs do not improve access. D:Missing objects should not be counted as successfully ingested.
Question 629 A source file includes a checksum in its manifest. Before parsing, what verification is most useful? Check only that the CSV has a header Validate only the file name and byte count without comparing the supplied checksum Compute or verify the delivered object checksum against the manifest Trust the object's last-modified time as proof that its contents match the manifest Explanation Correct Answer C:A matching checksum helps detect corruption or partial transfer of source bytes. Incorrect Answers A:A syntactically valid header does not prove the file bytes were delivered intact. B:Filename and size checks can miss content corruption; an end-to-end manifest checksum comparison verifies the delivered bytes. D:Modification timestamps do not establish content integrity or match the publisher's checksum.
Question 630 A producer delivers a data file twice under different names with identical content. The contract requires keeping both deliveries in the audit log but counting the business events once. What pattern fits? Discard both files and forget their arrival Aggregate by arrival hour only Retain delivery provenance and apply event-level deduplication under a stable source identity Use filename alone as the unique business-event key Explanation Correct Answer C:Operational delivery logs and logical event identity serve distinct purposes and can be preserved independently. Incorrect Answers A:This loses valid events and delivery evidence. B:Hourly aggregation cannot identify identical events among valid distinct data. D:Different names would defeat filename-only duplicate protection.