Migration strategy
How AuraShop moves from Abacus.AI to Microsoft Azure New Zealand North over four weeks, with a parallel run, two gates and a way back.
1.Hybrid model
Keep authoring AuraShop in Abacus.AI, and release it to Azure through a pipeline. Abacus.AI has let a small team build and ship a real product quickly, and it stays our place to make changes; Azure gives AuraShop its own home in New Zealand North to run in.
2.Four weeks, end to end
The week-by-week plan in summary. The full detail, stage by stage, is on the Timeline page.
| Week | Dates | What happens |
|---|---|---|
| Week 1 | 2–6 November | Stage 0. Decision and baseline; Stage 1. Code export and repo |
| Week 2 | 9–13 November | Stage 2. Azure landing zone; Stage 3. Data dry-run |
| Week 3 | 16–20 November | Stage 4. Release pipeline; Stage 5. Parallel run begins |
| Week 4 | by 27 November | Stage 5. Parallel run; Stage 6. Cutover |
Steady-state hybrid running, measuring and reporting follows from 30 November.
3.Two gates
Ready to build
End of October, owner sign-off.
Production cutover
Week 4, owner approval in the pipeline.
4.Parallel run and rollback
AuraShop runs on Azure alongside Abacus.AI before shoppers are switched. If anything goes wrong at cutover, DNS is pointed back to Abacus.AI, which stays available until steady state is confirmed.
5.Main risk
If the code can’t be exported from Abacus.AI, Stage 1 becomes a rebuild from the specification, and the dates are reviewed and recorded in the decision log.
6.How releases work
Changes are still made by describing them to the Abacus.AI agent. Each release is exported from Abacus.AI to the repository, and the pipeline builds, checks and deploys it to Azure, with owner approval for production.