LuminousMonkey

The Man Who Sold The World

I have a problem, I'm developing an application that uses a database (that's not the problem, databases are great, people should stop using them as stupid key/value stores), the problem I have is that I want to get continuous integration set up pretty early into the project. The earlier I get it organised, the better habits will flow from that (or that's the plan at least), and also the quicker it will be in the long run since it will help with the path to continuous deployment.

The problem I have, is that my CI server (TeamCity) is running under a virtual machine. Ok, that's not the problem exactly, but the problem is I want to have the state of my database in a known good state. Also, we need to make sure that it's documented how to build the system. For local development, I'm using Vagrant, but, since the CI server is running in a virtual machine itself, then Vagrant isn't really suitable. (Yes, you could run a VM in a VM, but lets just not.)

So, I either have to get a physical server running Linux (either migrate TeamCity to it, or have an agent run on it), or, figure out how to get a virtual machine that I can bring up and down via scripting. Luckily for me, it looks like someone already has:

https://dor.ky/using-vmware-vsphereesxi-as-a-provider-for-vagrant/

The basic idea is to create a template on your ESX server, then use a plugin for bringing a machine up and down based on that template. This seems like something that could work. The reason I need to bring up a whole database is because I'm not using some sort of ORM to abstract away the database. There are a few reasons for this:

  1. I would lose the use of triggers, which are important for maintaining data integrity, also they save the round-trips you can get with ORM.
  2. I can use the database authentication system, there's no point rolling my own when the database can do it better and faster.
  3. I don't need to switch databases, and designing the system for that means you end up coding the for lowest common denominator.

I have read some other blogs commenting both sides of the ORM debate, and I should write a post on it in the future about it, motivation willing. Anyway, I haven't got the build system up and running just yet, but I'm hoping that I've made some progress by Friday so should be able to do a quick post on how that's working out.