I'm running into an issue with not being able to create new migrations (with

prisma migrate dev

) after squashing migrations. Here's what I'm doing: 1. Squash migrations (following Prisma's guide on squashing migrations to create a clean history in a production environment: https://www.prisma.io/docs/guides/database/developing-with-prisma-migrate/squashing-migrations#how-to-create-a-clean-his[…]-a-production-environment), including marking the migration as applied 2. Make a change to

schema.prisma

3. Run

prisma migrate dev

to create a new migration 4. Make another change to

schema.prisma

5. Run

prisma migrate dev

to create another new migration 6. Migration fails with following error:

Error: P3006

Migration `20220718195944_migration_name` failed to apply cleanly to the shadow database. 
Error code: P1014
Error:
The underlying table for model `users` does not exist.

The same thing will happen if I run the first migration generation after squashing with `--create-only`: when I then subsequently try to run

prisma migrate dev

to apply the created migration, the same error message as above will come up. Creating/applying these migrations works fine as long as the previous migrations haven't been squashed. This is only happening after squashing. So I understand that I can revert back to the pre-squashed migration history and reset from there, but if squashing migrations is an option - and it's one we want to use - it shouldn't block further migrations from being created and applied. Any help/advice is greatly appreciated, thank you! UPDATE: This only happens when following the documented instructions for creating a clean history in a production environment, ie. by creating the

squashed_migrations

directory with the following command

npx prisma migrate diff \
 --from-empty \
 --to-schema-datamodel ./prisma/schema.prisma \
 --script > ./prisma/migrations/squashed_migrations/migration.sql

rather than creating it by running

prisma migrate dev --name squashed_migrations

.

Hi Vanessa, thanks for reporting the problem! I am not sure I understand all the moving parts in the problem. To recap, the error is only happening in development (with

migrate dev

), after you squashed your migrations and accepted a reset from migrate dev, right? One thing to keep in mind is that the migrations in

./prisma/migrations

are executed in lexicographic order, so if you keep the default generated names, any migration in a directory starting with a number (the timestamps) will come before

squashed_...

, so I think what you may be seeing is your migrations running in the wrong order (that would explain why your next migration fails on a missing table). Does that sound possible? Regardless, we have to fix the docs to use something like

000000_squashed_migrations

instead of

squashed_migrations

as a directory name.