Infrastructure as Code, 2nd Edition

Chapter 11. Building Servers as Code

Infrastructure as Code first emerged as a way to configure servers. Systems administrators wrote shell, batch, and Perl scripts. CFEngine pioneered the use of declarative, idempotent DSLs for installing packages and managing configuration files on servers, and Puppet and Chef followed. These tools assumed you were starting with an existing server—often a physical server in a rack, sometimes a virtual machine using VMware, and later on, cloud instances.

Now we either focus on infrastructure stacks in which servers are only one part or else we work with container clusters where servers are an underlying detail.

But servers are still an essential part of most application runtime environments. Most systems that have been around for more than a few years run at least some applications on servers. And even teams running clusters usually need to build and run servers for the host nodes.

Servers are more complex than other types of infrastructure, like networking and storage. They have more moving parts and variation, so most systems teams still spend quite a bit of their time immersed in operating systems, packages, and configuration files.

This chapter explains approaches to building and managing server configuration as code. It starts with the contents of servers (what needs to be configured) and the server life cycle (when configuration activities happen). It then moves on to a view of server configuration code and tools. The central content of ...