Skip to content
Accidental DELETE on production Postgres, no backups available: is a volume snapshot from before the deletion possible?
arangurenivan8-beep
OP

19 days ago

Problem and desired outcome

I accidentally ran a DELETE FROM "user" against our production Postgres database (I meant to run it against staging). The user row was removed, and because our foreign keys use ON DELETE CASCADE, everything tied to that account was deleted with it, including a large amount of shared-expense data that other users of our app also relied on. The Backups tab for the volume shows "No Backups" and we are on a plan without scheduled backups or point-in-time recovery. I would like to know whether Railway has any platform-level snapshot of this volume from before the deletion that could be restored (to a new volume, not over the current one), and if so how to request it. The data loss is real and the impact is ongoing for the affected users.

Resources and timing

Workspace / project / environment:

Affected service: the production Postgres service (select it in the support form):

Deletion time: <fill in date, time and timezone> on 2026-11-01. The statement was a single DELETE ... WHERE email = '' and it reported rowCount: 1.

Last known good state: the row existed minutes earlier; a SELECT immediately before the delete returned it (created on 2026-09-18).

Evidence and attempts

The delete was run with node + pg using the production DATABASE_URL, which I had copied by mistake. It returned borradas: 1 (1 row deleted) with no error, and a follow-up SELECT for that email returns an empty result.

The Backups tab for the production volume says: "This service's volume does not have any backups." and that creating backups or enabling point-in-time recovery requires Pro.

I have not changed anything else on the database since. I have not restored, vacuumed manually, or modified autovacuum settings. I have not upgraded the plan, because I understand that would not help for data deleted before backups existed.

Why contact support?

The AI assistant I was working with suggested contacting Railway support. Its reason: with no backups and no PITR, the only possible route to recover the data is a platform-level snapshot or disk-level copy that only Railway can access, and we have no access to the underlying volume or to Postgres extensions for reading dead tuples.

What I'm asking

Does Railway keep any snapshot or copy of this volume taken before the deletion, even outside the Backups tab? If so, can it be restored to a new volume or service so we can copy rows back selectively?

If not, is there any supported way to read the deleted rows from the current volume before autovacuum reclaims them, or to pause autovacuum safely in the meantime?

Constraints: please do not modify or restart the current production database or overwrite its volume. Everything else in production is live and in use.

Prepared with Railway's support template (v1).

Solved

1 Reply

Railway
BOT

19 days ago

Automatic snapshots, WAL recovery, and PITR aren't available to restore your data unless you've enabled them on the service. Volume backups and point-in-time recovery are both native Railway features on Pro. We don't recover data lost to user-initiated actions. Going forward, enable volume backups on your stateful services so you can self-restore if this happens again.


Status changed to Awaiting User Response Railway • 19 days ago


Status changed to Solved arangurenivan8-beep • 19 days ago


Welcome!

Sign in to your Railway account to join the conversation.

Loading...