Supabase deleting almost all my rows in 3 database tables

[fantasyproinsight]

deleting all the rows in a couple tables from 590 rows to 17 players and then completely wipe out records in another table. I don't see any logs .. How is this possible? I don't have code to delete from these tables? This behavior started to occur when I moved code to Vercel. I have tried this with a variety of database connection url strings. The code is Next.js. DATABASE_URL="postgresql://postgres:password@db.string.supabase.co:5432/postgres"
DATABASE_URL="postgresql://postgres:password@dbstring.supabase.co:6543/postgres?pgbouncer=true](http://db.pocqvmoyjhnpzldtefyl.supabase.co:6543/postgres?pgbouncer=true)"
Database URL currently in Vercel
postgresql://postgres.dbstring:password@aws-0-us-west-1.pooler.supabase.com:6543/postgres . I'm mystified. Has this every happened to anyone? Any possible reasons? Thanks

[GaryAustin1]

Supabase/Postgres does not delete data rows or tables.

You could check the API gateway logs for DELETE.
There is no log for database SQL thru the ports though.

Do you have anyone else with access to your Dashboard?
Have you ever used your service_role key?
Do you have the tables protected with RLS?
Is the data missing if you use the dashboard UI with the Postgres role?
Are you Pro with backups?

[fantasyproinsight]

every single connection string I try doesn't work in my deployed app with vercel. Sometimes it works sometimes it doesn't . I've tried pooled, direct, pgbouncer. It worked fine locally but doesn't like the environment variable setup in Vercel on redploy. And then the stupid cursor script would re-seed the database and remove a bunc of rows. Thanks if you have time for any suggestions

[fantasyproinsight]

Maybe I should have used v0 instead of cursor as then I'd have less migration issues from local to vercel

[fantasyproinsight]

it seems like vercel serverless has problems connecting to Supabase. It worked fine when running on localhost:3000 but since deployig to Vercel it's had all sorts of problems. MNot sure if it helps but hese are the errors I'm getting . Error fetching league settings: PrismaClientUnknownRequestError: Invalid prisma.user.findUnique() invocation: Error occurred during query execution: ConnectorError(ConnectorError { user_facing_error: None, kind: QueryError(PostgresError { code: "42P05", message: "prepared statement "s0" already exists", severity: "ERROR", detail: None, column: None, hint: None }), transient: false }) at $n.handleRequestError (/var/task/node_modules/.pnpm/@prisma+client@5.22.0_prisma@5.22.0/node_modules/@prisma/client/runtime/library.js:121:7505) at $n.handleAndLogRequestError (/var/task/node_modules/.pnpm/@prisma+client@5.22.0_prisma@5.22.0/node_modules/@prisma/client/runtime/library.js:121:6623) at $n.request (

[GaryAustin1]

I don't know much about Prisma but make sure you are running the right parameters including the pg_bouncer=true. This has to do with supavisor not supporting prepared statements.

https://supabase.com/docs/guides/database/prisma

[fantasyproinsight]

you are a tremendous help. That's all I needed was adding pg_bouncer to the end of the pooling string. Cursor was going around in a loop. I guess I should read the documentation better

[KenyStev]

It has just happened to our project today, out of a sudden most of the rows on several tables has been wiped out and our customer and raising issues for it, but the thing we haven't changed a thing in the db connection in a long time and this is the first time happening

[KenyStev]

I saw SOME rows in the tables but certainly not all of them, and in dashboard UI when I filter for specific Id it doesn't appears

[GaryAustin1]

The AI is correct on other possible ways than your code. All the other things are controlled by you including migrations.

[GaryAustin1]

Do you use the REST API?
Do you have delete RLS on the table to protect it?
Do you have more than you with access to the dashboard?
Someone with CLI access to your project?

You also seem to be saying it is 1 customer's rows that disappeared. That seems like a targeted delete from somewhere.
Just one table or cascades from one table?

[KenyStev]

Do you use the REST API?

• No I use the JS client

Do you have delete RLS on the table to protect it?

• No we haven't touched that

Do you have more than you with access to the dashboard?

• Yes, we are 2 having access but non of us has touched it recently

Someone with CLI access to your project?

• we both but haven't used it either and there's no logs related deletions either

no, sorry I mean I was testing with one customer id, but it was several for not to say all rows that vanished for all the customers

[GaryAustin1]

JS client IS the REST API. So you must have RLS protection. Do you have RLS enabled on the table? If so you must have a delete/select policy to protect from anyone deleting.

Check the dashboard REST API logs for Deletes.

Was it 1 user's rows only? If so seems something coming thru for that user IF you have RLS enabled and a delete policy.