if you need to drop an existing "production" database
RAILS_ENV=production DISABLE_DATABASE_ENVIRONMENT_CHECK=1 bin/rails db:drop
Set up new "production" database with 3 Foos and 3 Bars
RAILS_ENV=production bin/rails db:create db:migrate db:seed`
Precompile assets a la prod
RAILS_ENV=production bin/rails assets:precompile
Start a "production" server
RAILS_ENV=production bin/rails s
Go to
Both should load correctly
To simulate dropping a column during a production migration, keep the rails server running and run psql
RAILS_ENV=production bin/rails dbconsole
schema_cache_test_production=# ALTER TABLE foos DROP COLUMN x;
Now reload
/foos will fail because it makes a direct reference to the dropped column x
however /bars, which still does a Foo.all will succeed. But if you look at the column names, x will still be there since the application hasn't restarted to pick up the schema changes from postgres.
This is because Foo.all is doing a select foos.* (since enumerate_columns_in_select_statements = false)
Checkout the enumerate-columns branch to see the difference in behavior with enumerate_columns_in_select_statements = true