Check Your Knowledge

Test your understanding of this section’s capstone simulation before moving on to the glossary.

In the capstone scenario, what does the workload inventory’s classification of 500 stored procedures track back to? A. A random sampling of the codebase B. The Lift, Redesign, or Retire verdicts established in the autopsy section C. A vendor-supplied migration report D. The original Udemy course outline
On the go/no-go decision matrix, what does it mean to treat each criterion as a “gate” rather than a “score”? A. Criteria are averaged, and an overall percentage above 80% authorizes cutover B. Only the reconciliation criterion matters; the rest are informational C. A single non-green criterion holds the cutover regardless of how green the other criteria are D. Gates apply only to the first migration wave, not subsequent ones
In the war room scenario, what pattern in the row-by-row drill results pointed to a rounding-mode mismatch rather than a logic bug? A. Every mismatch was a different magnitude with no common pattern B. Every flagged row had a delta of exactly 0.01 or -0.01 on a value with an exact .5 fractional cent C. The mismatches only appeared on non-numeric columns D. The mismatches appeared on every row in the table, not a subset
What specifically caused the rounding divergence between Oracle and the migrated PySpark logic? A. Oracle’s ROUND function was deprecated and had no replacement B. Databricks does not support decimal precision beyond two places C. The migrated code used Python’s built-in round(), which ties to even, instead of pyspark.sql.functions.round, which is HALF_UP D. The reconciliation job itself introduced the rounding error
After fixing the rounding logic in the war room, what step closes the gap on rows already written during the mismatch window? A. Nothing further is needed once the function is corrected B. A full re-ingestion of the entire table’s history C. An idempotent MERGE INTO delta sync targeting just the affected rows, followed by a hash-diff check D. Manually editing the flagged rows in a spreadsheet
In “The Rollback Panic,” what are the two failure modes described as equally damaging? A. Rolling back too early and rolling back too late by the same margin B. Freezing on a decision past the checkpoint window, and overreacting with an unrehearsed rollback before diagnosis finishes C. Escalating to the steering committee and not escalating at all D. Running the reconciliation job twice and running it zero times
Why does rehearsing the rollback procedure in advance reduce “rollback panic” during a live incident? A. It eliminates the possibility of any future incident B. It removes the emotional judgment call from the moment of highest pressure by establishing a known cost and a known threshold beforehand C. It guarantees that a rollback will never actually be needed D. It shifts the decision entirely away from the steering committee
Which four artifacts does the capstone deliverable ask you to assemble from across this part of the course? A. A Gantt chart, a budget spreadsheet, a org chart, and a risk log B. A workload inventory with 3-R verdicts, a TCO comparison, a reconciliation plan, and a go/no-go matrix with a rollback plan C. A Delta Lake certification, a Unity Catalog audit, a Lakeflow pipeline, and a dashboard D. A vendor contract, a licensing audit, a hiring plan, and a training schedule
According to the course wrap, what is the main difference in what Part 2 (StepRight) and Part 3 (migration) each asked you to prove? A. Part 2 required more advanced SQL than Part 3 B. Part 2 built something new with no legacy baseline; Part 3 required proving a replacement is at least as correct as what it replaces C. Part 2 covered governance while Part 3 covered only ingestion D. Part 2 and Part 3 covered identical material with different table names
What does the course wrap identify as the more common shape of real-world Databricks work? A. Green-field builds with no legacy system involved B. Migrations and the trust-building work underneath them C. Reporting-only projects with no data engineering D. Single-user proof-of-concept demos

Answer Key

B – The inventory’s verdicts, established in the autopsy section, are what the 3-R decision and TCO calculator build on for the rest of the migration.
C – A go/no-go matrix is a set of gates, not a weighted score; one red criterion holds the cutover no matter how green the rest are.
B – A delta of exactly a cent on values with an exact .5 fractional component is the fingerprint of a rounding-tie disagreement, not a random data error.
C – Python’s built-in round() uses round-half-to-even on exact ties, while Oracle’s ROUND and Spark SQL’s ROUND (via pyspark.sql.functions.round) use round-half-away-from-zero.
C – The rows written before the fix still need an idempotent MERGE INTO backfill against just the gap, verified with the same hash-diff check used everywhere else in the reconciliation stack.
B – Freezing past the checkpoint window and overreacting before diagnosis completes are both failures of treating the rollback as an emotional call instead of a rehearsed procedure.
B – Rehearsal establishes the cost and the trigger threshold ahead of time, so the decision under pressure is executing a known procedure, not inventing one.
B – The workload inventory/3-R verdict, TCO comparison, reconciliation plan, and go/no-go matrix with rollback plan are the four deliverables the capstone asks you to assemble.
B – StepRight had no legacy baseline to match; this part required proving row-level and dollar-level parity against an existing system before it could be replaced.
B – The course wrap identifies migrations, and the reconciliation and rollback discipline underneath them, as the more common real-world case than green-field builds.

| ← Previous: Your Capstone Deliverable and Course Wrap | Next: Databricks Glossary → | |:—|—:|