Let's Get Scary and Continuously Deploy Database Changes!

Continuous deployment of our applications is easy! To roll back, all we have to do is just redeploy our old code and everything works again. Our application developers have been practicing these methods for years (even before the advent of automated CI/CD). But...databases are much scarier - how do we undo data changes? What if we dropped (or shrank) a column? How do we handle downtime? Do we actually trust our application developers? In this talk, we will discuss how to successfully automate database deploys as part of a CI/CD pipeline. We will address some of the stickier questions above, like how to handle data type changes and how to remove columns without losing access to the data. We will discuss automated testing in brief, and how it can help us avoid some of the messier problems. And lastly, we will showcase how to roll forward using both state-based and migration-based tools like SSDT or Liquibase. We will use Azure for most examples, and the concepts will apply to (almost) any platform.