What BackupDrill protects
You should know what your backups leave out before you need them. This page lists what a BackupDrill backup contains, what a restore drill checks, and what is not covered at all.
The coverage matrix
| Layer | Backed up | Restore-drilled | Notes |
|---|---|---|---|
| Database: public schema (tables, data, indexes, constraints, triggers) | Yes | Yes | In every backup. Drills restore it into a throwaway Postgres and check it against the manifest. |
| Database: additional schemas you own (CLI only) | If configured | If configured | Hosted projects back up public today. The open-source CLI can dump additional schemas you own via its schemas setting. |
| API grants for anon, authenticated and service_role | Yes | Yes | Recorded in each snapshot's manifest (CLI 2.9.0 and later). Restores and drills give these roles exactly what the source granted, whatever the target project grants by default, then read the grants back to check. Grants to other roles are not recorded. |
| Supabase Auth users and sign-in identities (auth.users, auth.identities) | Yes | Yes | User IDs, emails and phone numbers, password hashes, user metadata, and each user's Google, GitHub or email sign-ins. Saved as a second archive, auth.dump, next to the database dump. A restore writes them in the same transaction as the database. A drill restores them into the sandbox and checks the counts; if you have triggers on the Auth tables, it also inserts a test user straight into auth.users, the way an email sign-up writes one, and checks that the insert gets through those triggers without an error. On by default; you can turn it off per project. If the database role can't read the Auth tables, the snapshot records that and the database backup still runs. The console shows which snapshots have an auth.dump. |
| Auth sessions, MFA factors and SSO | No | No | After a restore everyone signs in again, and users who had MFA enroll again. |
| Storage files (uploads, avatars, documents) | If configured | If configured | Optional: add the project's Storage S3 credentials on its project page. Once added, files are synced on every backup with a sha256 for each in the manifest, and drills verify a sample of checksums. |
| Storage re-upload on restore | If configured | No | CLI restore uploads files back into the target's Storage (you supply the service-role key at restore time), checks every key is there, and spot-checks checksums. Console recovery restores the database only and gives you the equivalent CLI command. |
| Edge Functions source, Auth settings, PostgREST config, secrets | No | No | Not covered. SimpleBackups covers more of this config layer; see “Covering the gaps” below. |
The Restore-drilled column applies when your plan includes drills. Paid plans run them daily. The Free plan gets one drill on its first backup, then keeps backing up without scheduled drills.
The Auth rows need a closer look. User accounts live in the Supabase-managed auth schema, not in the schemas you own. BackupDrill exports the two tables that hold accounts and sign-in identities. When that export succeeds, they come from the same database snapshot as your tables, so they match each other. The rest of that schema (sessions, refresh tokens, MFA factors, SSO) is not exported, and Auth settings such as sign-in providers, SMTP and redirect URLs are not in the database at all. Supabase’s built-in backups cover the whole schema, so keep them on and use BackupDrill alongside them, not instead of them.
How the console shows your protection level
Protection has two separate parts, and the console shows both on every project, for example DB + Storage · daily drills or DB-only · one-time drill on Free. Aim for DB + Storage, drilled: your database and Storage files are in your bucket, drills test the database restore, and they check a sample of file checksums.
What is covered: DB-only vs DB + Storage
DB-only means your database schemas are in your bucket but Storage files are not. A database-only restore does not restore Storage files: files deleted since the backup stay missing, and in a new project none of the files are there. DB + Storage means both are in your bucket together, with a sha256 checksum for every file in the manifest.
Move up: Add your project's Storage S3 credentials (four fields) in the project settings.
Whether restores are tested: how often drills run
Drills restore your latest snapshot into a throwaway Postgres and verify it. If a drill fails, you get an email. How often they run depends on your plan: one drill on your first backup on Free, daily on every paid plan.
Move up: Upgrade to any paid plan for a restore drill every day. Free runs one, on the first backup.
What a passed drill proves, and what it doesn’t
A passed drill means your latest snapshot was downloaded from your bucket, its checksum matched, pg_restore completed into a clean Postgres, your indexes/constraints/triggers came back, the table count matched the manifest, populated tables restored non-empty, and row counts matched the dump exactly for ordinary tables and partitions. Materialized views, partitioned parents and tables owned by an extension get only the non-empty check. When the snapshot contains Storage files, a sample of their checksums matched too. When the snapshot has Auth users, the report adds an Auth line: whether they restored, whether the counts matched, and, if you have triggers on the Auth tables, whether they ran on a test user the drill inserts the way an email sign-up does (straight into the table, not through Supabase Auth). That line does not change pass or fail. That is strong evidence the backup can be restored. It is not a guarantee about every future restore, and it does not test Supabase-managed objects, which do not exist in the drill sandbox. The full check list is in the restore-testing guide.
Covering the gaps
- Auth settings and sessions: write down your sign-in providers, SMTP settings and redirect URLs somewhere outside the project, and keep Supabase’s built-in backups (or PITR) on for the parts of the
authschema we don’t export. - Storage files: add your Storage S3 credentials (the four fields in the project form). Without them, every file URL 404s after a database-only restore.
- Edge Functions and platform config: keep them in git; for a managed option, SimpleBackups covers more of that layer.
Not sure what level your project is at? Sign in to check. Every project shows its protection level next to its health.