Protocol for database updates

As of right now, there is no protocol for direct database updates.

1. Currently, the workflow that I perform is: change the database schedule that is checked in the repo.
2. Apply it by resetting the test server (test server automatically re-instals the latest version of this db schedule)
3. Rerun java / python unit tests
4. If nothing has changed to the API interface, apply this change somehow live.

This works OK-ish, but there are some flaws.
a) Live database is full of important information, so we can never reinstall the database. Updates have always to be performed on a running database (don't think we can change that)
b) As such, there is little guarantee that the database that is running on the test server is structure wise the same as the live database.

It would be great if someone could explore automated tools that compare a running database against a sql schema, and indicate what the differences are (this would solve point b). It would be even better if someone would find tools that could update a live running database with a schema that is provided (auto fix a database against a schema. This would solve point a)

Start with downloads/openml_expdb.sql and the described test-server reset workflow. Investigate tools that compare a running database with this schema and can safely update a live database, then rerun the Java and Python unit tests. Done means the workflow detects schema differences and provides a reliable path to apply compatible updates without reinstalling the live database.