## Summary

Accidentally ran prisma migrate reset --force --skip-seed against my production PostgreSQL database while configuring a separate demo environment. I copied what I believed was the demo DATABASE_PUBLIC_URL but it actually pointed to production. All real customer data was lost.

This is a real production system used by a paying client (a salon business in Peru). I am NOT asking Railway to perform recovery on my behalf. I just need a definitive yes/no on whether ANY infra-level recovery is technically possible for this volume before I declare the data permanently lost and move on to external-source reconstruction.

## Specific questions

1. Are there any volume snapshots retained at the infrastructure level (even non-user-facing) for this Postgres service?

2. Are WAL segments still on disk that could enable a point-in-time recovery to a moment before the reset?

3. Any internal block-level backup retained for the volume?

If the answer to all three is no, please confirm so I can close this and proceed with external reconstruction (SUNAT records, Cloudinary assets, payment processor history, etc.).

## Timeline (UTC)

- 2026-04-29 ~23:00 UTC - Created a new "demo" project + new Postgres service

- 2026-04-29 ~23:15 UTC - Copied what I thought was the demo DATABASE_PUBLIC_URL (actually production)

- 2026-04-29 ~23:20 UTC - Ran prisma migrate reset --force --skip-seed

- 2026-04-29 ~23:25 UTC - Ran prisma db push --accept-data-loss

- 2026-04-29 ~23:30 UTC - Ran node prisma/seed.demo.js (wrote demo data on top of the wiped schema)

- After that I paused writes to the database

## Project info

- Production Postgres host: metro.proxy.rlwy.net:22323

- Database name: railway

- ORM: Prisma 5.7.0

- Plan: Hobby

- I did NOT have the volume Backups feature scheduled (lesson learned, will be enabling immediately across all my projects)

If you need exact Project ID and Service ID I can provide them - just let me know.

## What I have already done

- Stopped writes to the database (paused backend service)

- Documented full timeline

- Identified external sources for partial reconstruction

- Reviewed local git history and disk for any local pg_dump (none found)

## What was lost

Customer records, all appointments, SUNAT receipt DB references (receipts themselves safe at SUNAT), WhatsApp campaign data, Cloudinary image references (files themselves safe in Cloudinary), and owner-customized site settings.

Thank you for any clarity you can provide. Even a definitive "no" lets me move forward with reconstruction work.