Wave cutover and rollback

Contoso's schedule is fixed by its strategy course. The first three months after the S/4HANA go-live are stabilization and hypercare, with no platform moves. Migration waves run from go-live + 3 to go-live + 12 months, starting with a four-interface pilot, and every one of the 57 configurations not retired leaves PO by go-live + 12 months, well before PO 7.5 mainstream maintenance ends at the end of 2027. This lesson is the mechanics of one wave: transports, the per-interface switch, the rollback, and the evidence that closes it.

One wave, end to end

1. Build. Migrate or build the wave's flows in the CMC_DEV tenant; run layer 1 and 2 tests there.
2. Transport to test. CMC_TEST imports automatically. Run the end-to-end replays; Priya reviews the difference reports.
3. Sign off. Priya forwards the transport from CMC_TEST only after test sign-off; the business signs the wave (Dan for finance, Maria as O2C key user for order to cash).
4. Import to production. CMC_PROD imports on its schedule in the approved release window; Luis, as release manager, runs it. Flows are deployed but receive nothing yet.
5. Switch each interface with its runbook (below), one at a time, watching the first real messages.
6. Hypercare. PO configuration stays intact and stopped; message counts are reconciled daily until the wave closes.

The per-interface switch

Take Wave 1's outbound purchase order IDoc to Lakeside Components. Today CMP sends the ORDERS IDoc through a tRFC port to PO, which maps it and writes a file to Lakeside's SFTP server. After the switch, CMP sends to the Cloud Integration IDoc endpoint and the new flow writes the file. The runbook: confirm Lakeside has allowed the tenant's IP addresses; stop the PO sender channel in communication channel monitoring; let in-flight messages drain and record PO's counts; change the outbound parameters in the partner profile (WE20) to an XML HTTP port (WE21) whose SM59 HTTP destination points at the tenant; release one real order; check with Lakeside that the file arrived complete and readable.

Rollback is the reverse: switch the partner profile back to the PO port and restart the PO channel. It only works while the PO configuration is untouched, so nothing in PO is deleted until the wave's hypercare ends. Write rollback triggers down before the switch, for example: the first real message fails, the partner rejects a file, or counts don't reconcile at the end of day one. Whoever is on shift may roll back without a meeting; the decision is reviewed afterwards.

Retirements and the last wave

The seven retirements are not migrated. Their channels are stopped and their configuration exported, then kept for one full year-end cycle before deletion, because a zero-traffic interface may be a year-end or audit feed. That's why migration wave 4 confirms them after year-end. The last wave also leaves the PO decommission plan ready: no active channels, archived message data under its retention rules, and no other system still pointing at PO.