P3005: The database schema is not empty.

TL;DR: Your database already has tables but Prisma has no record of creating them. Tell Prisma the current schema matches your first migration with prisma migrate resolve --applied "[first-migration-name]", then deploy normally. Do not drop the database to fix this.

Error: P3005: The database schema is not empty.

Steps

1. Confirm the situation: tables exist, migration history does not:

   npx prisma migrate status

Expected: it reports P3005 or shows every migration as pending against a non-empty database.

1. Verify your first migration actually describes the existing schema. Diff it:

   npx prisma migrate diff --from-migrations ./prisma/migrations --to-schema-datamodel prisma/schema.prisma --shadow-database-url "$SHADOW_DATABASE_URL"

Expected: an empty diff. If the diff is not empty, fix the schema first; baselining a mismatch just moves the pain.

1. Baseline the database:

   npx prisma migrate resolve --applied "20240101000000_init"

Use the real name of your first migration folder. Expected: Migration 20240101000000_init marked as applied.

1. Deploy:

   npx prisma migrate deploy

Expected: No pending migrations to apply or only genuinely new migrations apply.

When to use

• The exact code is P3005 on migrate deploy against a database built by db push, a dump restore, or another tool
• Adopting Prisma Migrate on a brownfield database

When not to use

• The database is disposable (local dev): prisma migrate reset is faster and honest
• You have real migration history and a migration failed halfway: that is P3018, not P3005

Compatibility

Prisma Migrate 3.x through 6.x. PostgreSQL, MySQL, SQL Server.

Variant phrasing: P3005 right after prisma db push in staging

db push creates tables without writing migration history. Baseline once, then use migrate deploy from then on.

Variant phrasing: P3005 on a restored production dump

The dump has tables but no _prisma_migrations rows. Baseline against the migration that matches the dumped schema.

Why it happens

migrate deploy refuses to run its CREATE TABLE statements against a non-empty database because that would destroy or duplicate existing tables. The --applied baseline writes a history row claiming the first migration already ran, so Prisma treats the existing schema as its starting point.

Edge cases

• Baseling the wrong migration name silently lies to Prisma; later migrations may fail in confusing ways. Verify with the diff in step 2.
• If the existing schema only partly matches, consider db pull into a fresh schema first to capture reality.
• Some teams baseline every environment once and document the command in their deploy runbook so nobody re-runs it.