Lesson 3 of 5
Rehearsals
11 min
A cutover you have not rehearsed is a cutover you are doing for the first time in production. Rehearsals, usually called mock cutovers, run the runbook end to end in a production-like environment, with production-like data volumes, the real team and the real clock. The final rehearsal, the dress rehearsal, proves that the final runbook works as written, in the time available, with the people who will do it on the night.
What each rehearsal is for
| Mock 1 | Mock 2 | Dress rehearsal | |
|---|---|---|---|
| Purpose | Find sequence, mapping and tooling gaps | Prove full volumes, timings and fallback | Prove the final plan exactly as it will run |
| Data | Full or large sample | Full production-like volumes | Full volumes, fresh copy |
| People | Core team | Real owners, including business validators | Exactly the go-live team and roster |
| Overruns | Expected and measured | Should be small and explained | Not acceptable; must finish without heroics |
| Exit | Every task attempted; issues logged | Fits the window; fallback timed | Fits the window with agreed contingency; no open go-live critical issues |
Contoso's rehearsals
Mock 1 took 71 hours against Contoso's 60-hour window. That was the useful result: it showed where the time went, most of it in waits between tasks and in a load that had to wait for a manual step. Mock 2, with a corrected runbook and full volumes, fitted the window with only 1.5 hours of float, and it also measured the fallback: restoring the pre-cutover backup and reopening the legacy system took nine hours. After Luis moved static checks into pre-cutover and parallelized stock reconciliation, the dress rehearsal finished with four hours of contingency.
- Late RealizeMock 171 hours against a 60-hour window; waits and sequence gaps found
- Early DeployMock 2Fits with 1.5 hours of float; fallback measured at nine hours
- BetweenRunbook changesStatic checks moved out of window; stock reconciliation split by plant
- Deploy, about three weeks before go-liveDress rehearsalFinishes with four hours of contingency; runbook frozen
Making a rehearsal realistic
- Environment. Use a production-sized copy, ideally built the way production will be, so load and conversion timings mean something.
- Data. Use a recent copy of production data, protected according to your data privacy rules, with full volumes by Mock 2.
- People. The people who will do the job on the night, including business validators and approvers, not stand-ins. A stand-in hides how long the real approver takes.
- Clock. Run in shifts across a weekend if that is how the real cutover runs. A rehearsal spread over two weeks of office hours proves nothing about timing.
- Late changes. Include every fix and transport planned for go-live. A dress rehearsal that excludes late changes does not prove the real plan.
Measure everything
Assign a timekeeper whose only job is to record actual start and end times, and any waits, for every task. After each rehearsal, compare actuals with plan, recalculate the critical path and the remaining float, and write a lessons log. Every lesson should lead to a runbook change or a recorded decision not to change.
- RunExecute the runbook against the clock
- RecordTimekeeper logs actual times and waits
- AnalyzeCritical path, float, failed and late tasks
- ChangeRunbook, work instructions, roster, data fixes
- Re-proveNext rehearsal shows whether the change worked
Test the fallback too
Rehearsals should prove not only the way forward but also the way back. Mock 2 is a good place to execute the fallback: restore the target to the pre-cutover backup, reopen the legacy system, re-enable interfaces and confirm that business can resume. Measuring how long that takes gives you the latest safe time to decide on fallback, which Lesson 5 uses.
What would you do?
The team proposes to skip the dress rehearsal because Mock 2 went well and the developers want more time for performance fixes. What should Grace decide?
- Mock cutovers run the runbook end to end with production-like data, real people and the real clock.
- Mock 1 finds gaps, Mock 2 proves volumes and fallback, the dress rehearsal proves the final plan.
- A timekeeper records actuals; every lesson becomes a runbook change or a recorded decision.
- Test the fallback in a rehearsal and measure how long it takes.
Check your understanding · not graded
Lesson check
1. What is the purpose of the dress rehearsal?
2. Why should rehearsals use the real business approvers rather than stand-ins?
3. What should happen to each lesson from a rehearsal?
4. Why test the fallback during a mock?