Check Your Knowledge

Test your understanding of sequencing a legacy migration into a single runbook, the material covered in this section.

1. What is the primary purpose of a migration runbook, as distinct from the individual design documents produced by earlier phases (workload inventory, schema translation, reconciliation)? A. It replaces the need for a risk register B. It stitches deliverables into one ordered sequence with owners, entry/exit criteria, and rollback instructions per step C. It is a marketing document for the steering committee D. It only covers the cutover phase, not the full migration

2. Why is “reconciliation passed” not considered an adequate exit criterion for a runbook step? A. Reconciliation should never be part of a runbook B. It isn’t specific or verifiable – a real exit criterion names the exact check and result required C. Exit criteria are only needed for the final step D. Reconciliation results can’t be logged

3. In the sequencing dependency graph, why is schema freeze treated as the root node? A. It’s simply the first step chronologically, with no dependency implications B. Nothing downstream can be trusted until the source schema it’s built against is fixed C. Schema freeze is only relevant to the risk register, not the sequence D. It has no bearing on DDL translation

4. Why is logic conversion (PL/SQL-to-PySpark transpilation) modeled as a hard block rather than something that can run in parallel with reconciliation for the same domain? A. Hard blocks are simply a formality with no technical basis B. Reconciliation checks the output of the converted logic, so there’s nothing to reconcile until conversion is complete and reviewed C. Reconciliation must always run before conversion D. Parallelism is never allowed anywhere in a migration sequence

5. At what level does ingestion and reconciliation typically parallelize in the sequencing model described? A. Per table, once conversion for the domain is complete B. Per column C. Only at the very end of the entire migration D. Never – reconciliation is always fully sequential

6. What is the single most valuable field in a risk register entry, according to this section? A. Risk description B. Probability C. Owner – the named person who monitors the risk and triggers the mitigation D. Impact

7. What distinguishes a steering committee from a general status meeting? A. It meets more frequently B. It includes someone with actual authority to approve a cutover or authorize a rollback C. It only discusses budget D. It excludes business sponsors

8. Why does trigger-based escalation exist as a separate communication cadence from scheduled weekly status updates? A. To reduce the total number of meetings B. So a risk crossing its mitigation trigger reaches decision-makers immediately, rather than waiting for the next scheduled update C. Because weekly updates are prohibited during active phases D. Trigger-based escalation replaces the risk register entirely

9. Why do most legacy-to-Databricks cutovers converge on a 72-hour rollback window rather than a shorter or longer one? A. It is a regulatory requirement in every jurisdiction B. It covers at least one full business-day cycle plus a buffer to catch issues that only surface on a specific day, without indefinitely prolonging a costly warm standby C. It matches the default retention period of Delta Lake’s transaction log D. It is exactly the duration of a shallow clone’s validity

10. What does “warm standby” mean for the legacy system during the rollback window? A. The legacy system is fully decommissioned but its backups are retained B. The legacy system actively serves read-write production traffic alongside Databricks C. The legacy system stays untouched and read-only, verified capable of resuming as system of record within its tested recovery time D. The legacy system is powered off but its VM image is kept

Answer Key

1. B – A runbook’s purpose is sequencing prior deliverables into one ordered plan with owners and entry/exit criteria per step, not producing a new deliverable itself.
2. B – “Passed” is not verifiable on its own; an exit criterion needs the specific check and the specific result required to advance.
3. B – Schema freeze is the root because DDL translation, code conversion, and reconciliation are all invalid if the schema they’re built against later changes.
4. B – Reconciliation is checking the output of converted logic, so it has nothing to check against until that conversion is complete.
5. A – Once a domain’s conversion is done, reconciliation and ingestion have no dependency between tables and can run concurrently per table.
6. C – Owner is what turns a named risk into someone actually watching a specific signal and triggering the mitigation.
7. B – A steering committee without someone who can approve a cutover or authorize a rollback is just a status meeting.
8. B – Trigger-based escalation exists to shorten the gap between a risk materializing and decision-makers hearing about it, rather than waiting for the next scheduled cadence.
9. B – 72 hours covers a full business cycle (batch jobs, day-of-week reports) with margin, without the cost of an indefinitely extended warm standby.
10. C – Warm standby means untouched, read-only, and verified ready to resume as system of record within the backfill procedure’s tested recovery time – not merely “still powered on.”

---

| ← Previous: The 72-Hour Rollback Window | Next: The War Room: Capstone Simulation → | |:—|—:|