Lesson 4 of 5
Rehearsal
12 min
Listen to this lesson
Rehearsals turn a plan into a measured, practiced procedure. For a conversion, each rehearsal is a full or partial conversion of a fresh copy of production, run from the runbook, by the people who will run the real weekend. SAP Activate places the final rehearsals in the deploy phase; the earlier test conversions start in realize. Contoso runs two mock conversions and a dress rehearsal.
What each rehearsal is for
| Mock 1 | Mock 2 | Dress rehearsal | |
|---|---|---|---|
| Main goal | Find gaps and measure | Fit the window; prove fallback | Run it exactly as production |
| Source | Fresh copy of EP1 | Fresh copy of EP1 | Fresh copy of EP1, as close to the weekend as possible |
| Runbook | Draft; many changes after | Corrected; timings tightened | Final; frozen afterwards |
| Validation | Method tried | Full reconciliation and verification | Full, with sign-offs against the clock |
| Contoso result | 71 hours: overran by 11 | 58.5 hours; fallback measured at nine hours | Finished with four hours of contingency |
Rules for useful rehearsals
- Production-sized data. Use a fresh copy of production each time; runtimes and data issues depend on volume and content.
- Same hardware class. Run on infrastructure comparable to the production target, or the times do not transfer.
- Same people, real shifts. Run across a weekend in shifts, as the real cutover will. A rehearsal spread over office hours proves nothing about timing or fatigue.
- Follow the runbook literally. Every deviation is logged and becomes a runbook change, not a memory.
- Validate fully. Reconciliation, verification tests and sign-offs are rehearsed too, with the same reports and variants as the weekend.
- Test the fallback. At least once, measure how long it takes to return to ECC.
Rehearsals also feed test management. Each mock yields a converted system that the test team uses for a regression cycle, and SAP's Data Transition Validation tool can be practiced on mocks so its comparisons are trusted on the weekend. Priya's team logs rehearsal defects in SAP Cloud ALM like any other, with the runbook step that exposed them.
Keep the copy as close to the weekend as you can for the dress rehearsal. Data grows between rehearsals, and new data brings new errors: a supplier created last month with a missing tax number, a document posted with an unusual combination of fields. A dress rehearsal on a copy taken a few weeks before go-live shows the volumes and data the production weekend will see, and leaves time to fix what it finds.
The runbook in SAP Cloud ALM
The runbook is the single list of every cutover step, and the rehearsals are how it gets right. Contoso keeps it in SAP Cloud ALM, the program's ALM tool: each step is a task in the project, assigned to an SAP Activate role and a named person, with its system, planned duration and predecessors. Milestones mark the decision points, the Gantt-chart view shows the dependencies and the critical path, and the transports for CMP are planned as a deployment in a release. During a rehearsal, shift leads record the actual start and end on each task as it happens, so timings come from the tool and not from memory.
| Step | What | Owner | Predecessor | Planned | Evidence |
|---|---|---|---|---|---|
| RB-0120 | Suspend released jobs with BTCTRNS1 | Sam | RB-0110 interfaces stopped | Fri 20:30, 0.5 h | Saved SM37 job list |
| RB-0300 | Confirm DOWNCONF_DTTRANS; SUM downtime starts | Sam | RB-0250 users locked, backup verified | Fri 23:00 | SUM log entry |
| RB-0610 | Finance reconciliation, company code 1710 | Dan | RB-0550 finance migration complete | Sun 10:00, 3 h | Signed reconciliation file |
| RB-0800 | Go/no-go 2 | Grace | RB-0610 to RB-0690 done | Sun 14:00 | Decision record |
Four of Contoso's runbook steps. Every step has an owner, a predecessor, a planned time and the evidence it leaves.
Measure, then change the runbook
- RunExecute the runbook on a fresh production copy, in shifts
- RecordPlanned and actual times, waits, errors and deviations for every step
- AnalyzeCritical path, longest waits, repeated errors, float used
- ChangeFix causes in ECC, move work out of the window, parallelize, update owners
- Re-planNew timeline and contingency target for the next rehearsal
Contoso's biggest gain came from waits, not from SUM. In Mock 1, many steps sat idle until someone noticed the previous step had finished. Luis added explicit handover messages and a shift lead per area. Finance migration errors that took four hours to analyze were fixed in ECC before Mock 2. Static checks moved before Friday, and stock reconciliation was split by plant. Those changes, not faster hardware, gave the dress rehearsal its four hours of contingency.
What would you do?
At 02:00 on Sunday of Mock 2, the 1710 finance reconciliation shows a difference. The shift lead finds a wrong entry, corrects it directly in CMQ, re-runs the report and marks the step done. What should Luis do in the morning?
- Rehearse on fresh production copies, comparable hardware, in real shifts, with the real people.
- Mock 1 finds gaps, Mock 2 fits the window and measures fallback, the dress rehearsal proves the final runbook.
- Record planned and actual times for every step; most savings come from waits and moved work.
- Freeze the runbook after the dress rehearsal; rehearse validation and sign-off, not just SUM.
- Keep the runbook as owned, dependent tasks in SAP Cloud ALM; record actual times and evidence on each step.
Check your understanding · not graded
Lesson check
1. Why must rehearsals use a fresh copy of production?
2. Which result marks Mock 2 as useful for fallback planning?
3. Where did Contoso find most of its time savings between rehearsals?
4. The dress rehearsal finished only because three people worked twenty hours straight. Is it a pass?
5. Where do the timings in Contoso's rehearsal analysis come from?