Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

Rehearsing cutover on a copy: what a dress rehearsal proves

Technical article
Hosting

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

A dress rehearsal is a full calendar walk-through of cutover steps against a non-production copy of the hosted stack and representative integrations. It answers whether named people can execute the runbook in the allotted window with realistic handoffs. It does not prove every production edge case, and it is not the same as a disaster-recovery exercise typology or a months-long parallel run.

Material difference from published coverage: What a tabletop, a partial restore and a full cutover each prove classifies exercise types for recovery and migration. This article is a timed cutover runbook rehearsal on a copy with explicit can/cannot prove lists.

Why it matters

Production cutovers fail on coordination: DNS owner asleep, MFA not tested, print step missing. Rehearsal exposes those gaps when rollback is cheap. NIST SP 800-84 (September 2006) is a guide to design and evaluate test, training and exercise programs for IT plans; migration dress rehearsal is inspired by that discipline, not a claim of SP 800-84 compliance or quoted timeboxing from the document.

For Australian businesses with weekend windows, rehearsal also reveals whether planned task durations fit the window — illustrative example: four hours of tasks might consume six when humans read checklists (not a measured statistic).

How it works

Scope rehearsal to cutover mechanics you control:

  • Final replicate or backup restore into a sandbox hosted environment. Before you start the copy, isolate it from production networks and identity (separate VLAN, no production domain join, no shared production service accounts). Disable or redirect outbound email, payment gateways and scheduled integrations so cloned jobs cannot modify production — a non-production label and privacy controls alone are insufficient. Production data in a copy needs the same privacy and access controls as production (customer responsibility). These steps have not been lab-tested.
  • Identity switches that can be simulated (test OU, test conditional access). Untested — your identity owner decides whether test OUs and policies are safe in your tenant.
  • DNS changes on a test zone or hosts-file simulation for app teams (untested template).
  • Execution of UAT must-pass tasks on the copy.
  • Communication tree dry-run (pager numbers, not only email).

What rehearsal can prove (editorial)

  • Runbook step order is complete and owners are named.
  • Estimated durations fit the chosen weekend or weeknight window.
  • Rollback procedure is executable, not theoretical.
  • Support bridges and vendor bridges connect.

What rehearsal cannot prove (editorial)

  • Production data quirks (historical corrupt rows, odd integrations).
  • Production DNS cache behaviour worldwide.
  • Peak Monday concurrency on session hosts.
  • Vendor escalation behaviour under live contract support paths (editorial — not a penalty simulation).

Use a single clock: start at the same notional time as production (for example "T-0 = Saturday 05:00 Sydney time") and record overruns.

What to choose and why

Editorial: choose one full rehearsal minimum for a first hosting cutover; skimp only when an identical pattern was rehearsed within months on the same stack.

Choose production-like access paths — if users will use browser client, do not rehearse with admin RDP only.

Choose stop criteria when rehearsal overruns window by more than agreed slack; fix runbook before repeating.

Record defects as runbook edits, not tribal knowledge.

Editorial view: one timed rehearsal often beats many untimed walkthroughs — your programme may differ.

Risks and limits

Copy drift. Sandboxes without refresh misrepresent integration versions.

False confidence. Green rehearsal with IT-only actors hides business-task failures.

Destructive steps. Production DNS or AD writes belong in production only; simulate safely.

Related: How it works.

Sources

Learning notes

  • Dress rehearsal tests people, timing and runbook — on a copy.
  • It does not replace production-only risks like DNS cache.
  • Use the same client access path as go-live.
  • Overruns mean fix the runbook, not hope for faster humans.
  • Treat identity and DNS simulation steps as untested until your team validates them.