3.4.1. Deployment strategy¶
A successful deployment strategy is highly connected with a good versioning strategy. To consistently run a successful develop, test, deploy cycle, it is important to be able to consistently version the full application stack you are running on each environment and to be able to recreate this across environments.
The modular architecture - as exposed through Helix - is not to be confused with a modular deployment model where a base version of the system is deployed and modules are installed on top of it – and where the product lifecycle of the modules, underlying system and websites on top of it are independent.
In Helix – as with development and Sitecore generally – it is highly recommended to follow a strict develop, test and deploy lifecycle for the whole application. This means that although development of the individual modules might have different development speeds, the assembly and versioning of the application stack happen at a particular point in the application lifecycle, before the deployment onto the environments. In other words, it is discouraged to do partial deploys of modules or features from development to production. Deployments should happen on an application-level scale, which fits with the availability requirements of the business.
Managing the full application stack - which includes vendor systems such as SQL Server and Sitecore XP, modules for those systems, and the underlying operating system - can be handled in a variety of ways. The best option often depends on the level of automation of the deployment process. In most cases this is done through documentation. It can also be handled by manifests that are read by automation systems and that then automatically pull in the right underlying systems to create the application stack in the build process. This methodology is done by popular package systems such as Node/NPM, NuGet, Bower and even the Sitecore Instance Manager. This methodology can also be employed as part of your custom build process to build more complete deployment packages.
Habitat leverages a number of package tools including NuGet, Node/npm and Bower to pull versioned packages into the build process.
The internal continuous integration and build server for the Habitat project – which is hooked up to the public GitHub repository – is set up to create versioned Sitecore packages on nightly builds or releases. This means that the full Habitat application with all modules can be installed as a single package onto any environment – and thereby it is possible to consistently recreate a given version.
Furthermore, the server fully automated reinstalls the Sitecore IIS website on an internal test server as part of the automated build. This includes reinstalling a blank Sitecore XP and required modules, and installing the Habitat content and application package on top. Even though this server is not publicly available as an example, parts of the build scripts, specifically the Gulp scripts, for the CI process are in the Habitat GitHub repository.
Tools such as Microsoft Team Foundation Server or Octopus Deploy can help immensely with automating of the build and deployment process and thereby in securing consistent versioning and deployment across environments.