Python for DevOps

Chapter 10. Infrastructure as Code

Before we had fancy DevOps titles and job descriptions, we were lowly system administrators, or sysadmins for short. Those were the dark, pre-cloud days when we had to load the trunks of our cars with bare-metal servers and drive to a colocation (colo) facility to rack the servers, wire them, attach a wheeled monitor/keyboard/mouse to them, and set them up one by one. Grig still shudders to think about the hours he spent in colos, in blinding light and freezing A/C. We had to be wizards at Bash scripting, then we graduated to Perl, and the more fortunate of us to Python. As the saying went, the internet circa 2004 was held together with duct tape and bubble gum.

Somewhere during the period of 2006 to 2007, we discovered the magical world of Amazon EC2 instances. We were able to provision servers through a simple point-and-click interface, or through command-line tools. No more driving to colocation facilities, no more stacking and wiring bare-metal servers. We could go wild and launch 10 EC2 instances at a time. Or even 20! Or even 100! The sky was the limit. However, we quickly figured out that manually connecting to each EC2 instance using SSH and then setting up our applications on every instance separately was not going to scale. It was fairly easy to provision the instances themselves. What was difficult was to install the required packages for our applications, add the correct users, make sure the file permissions looked right, and finally ...