Currently I have a Rails 2.3 app that is a CMS serving over 100 websites. To encapsulate things a bit better each site has it’s own copy of a standard database structure (as defined in the schema.rb file). I currently use MySQL and when a request comes in to the server I check the hostname, look this up in a table in a central database, and then switch the connection to the appropriate database.
I’m looking to rewrite this application in Rails 3.1 and am also investigating using sqlite3 in production instead of MySQL as the individual databases are fairly small and would suit sqlite3. I’ve also used it in production on another project and it works well.
I’m looking for both tips on how to switch sqlite3 databases in-app, plus also how to detect the currently connected database within Rails for testing purposes.
For those that are contemplating this. Turns out it is much easier to implement this using PostgreSQL schema’s. It’s called multi-tenanting and with postgres you can have one database (defined in database.yml) and multiple schema’s within that database which are essentially just namespaces. Each schema has its own tables etc… and you can create, switch between and delete them on the fly. Make sure you use at least Rails 3.1.1 as prepared statements apparently wreaks havoc with database and schema switching. There are also some fixes in Rails master as of writing that address some minor issues with schemas.
There is also a gem out there for multi-tenanting based on the above technology. It’s called Apartment but I just implemented things manually in my app as it was only about 20 lines of code to do what I wanted vs trusting a gem to do all the voodoo 🙂