I am going to talk about our Sitecore environment including the different servers involved, the process of building new functionality and fixing bugs. This post is the product of a lot of blood, sweat, tears + money. It hopefully might give you some direction in how to set up your environment. The project has gone from simple to complex, back to simple and probably ended up somewhere in between. Our set up is probably not in its final state yet as we have found that things don’t last long in our Agile environment. Constant tweaks here and there in aid of improvement.
To start with, we wanted a robust, tested system with as many regression tests as possible. To support this we chose to use continuous integration (CI). The annoying thing is that Sitecore and CI don’t really fit hand in glove. The main pain is the fact that a working environment is a mix of code and data. We found it nice and easy to add code into Subversion and for that code to be automatically checked out, built and tested on a CI box. However how do you handle data changes? For example you could delete a bunch of nodes in Sitecore and break the site, change the template types, change names, change permissions. How do you tie this in with the versioned code.
We decided that trying to directly tie data into the CI process would be super hard. You could have some event serialize the items in the tree and do something with them but we decided it was a time sink for the project.
Moving forward we have stuck with two real areas for testing for our CI, firstly using WatiN for the front end of the web site as well as unit testing and integration tests for core business logic.
Development Environment a.k.a. “Dev Dirty”
This is where the ball starts rolling. This is the developers local machine. The developers local machine is connected to a shared database, shared amongst the other developers. (Caveat – We try not to work on the same task at the same time as we find caching can cause an issue (usually resolved with an IE cache clear and IIS reset)).
We have 2 main branches in our Subversion repository. One called “current” which is the current code we are working on in our development environments and one called “test” for our test environment.
When we pull a feature task off the Scrum board, we would create a branch off the “current” and call it something like Feature – [Seb's New Feature]. This means I can go off and work on this new feature. When I am done with the code I create WatiN tests that test the frond end UI and any business logic testing. We use a Model View Presenter pattern to break open the larger/more complex stuff so that we can test it.
When I am done and happy, I will check this back into to my Feature – [Seb's New Feature] branch. I will then jump back on to the current branch and merge my feature back in. (I am told that this is the easiest way of doing it).
Cruise Control Environment
Upon me checking code back into the current branch, a seperate box running Cruise Control picks up on the fact that something in the code has changed in Subversion. It then runs a custom NAnt build file that pulls out the latest version of code from Subversion, compiles it, tests all our of Nunit and Watin tests and then alerts us to whether or not the build worked. You can at this point make it even more swanky and have it check code coverage (NCover), quality of the code (FXCop) etc…
Development Environment a.k.a. “Dev Clean”
We have experimented with pushing information between environments via packages as well as full blown system publishes (add the destination as a publishing target and it will overwrite it for you). We are currently trying to stick to piecemeal packages for each task/bug we work on. We can then commit those into Subversion under our branch so we can see which code was related to which piece of Sitecore Data.
To give an example, say I changed 2 templates in Sitecore as part of a new feature, I would create a package containing them, and save it under a folder called “Uninstalled Packages” under my feature branch.
This process gives us the ability to logically group and version code + data.
Any way back on track – The Development environment is the next environment upstream of our Local development machines. It is still a development environment, however this is the make or break point to test to see whether our packages work.
We will do an update on the codebase using subversion behind the scenes and install the package using the package installer in Sitecore. If it doesn’t work we go back and re-create the package. Repeat until it works, and re-commit to Subversion.
We are now ready to push onto a Quality Assurance testing environment. The environments upstream of Dev will be covered over the next posts
Tune back in for more.




