Book Image

HashiCorp Packer in Production

By : John Boero
Book Image

HashiCorp Packer in Production

By: John Boero

Overview of this book

Creating machine images can be time-consuming and error-prone when done manually. HashiCorp Packer enables you to automate this process by defining the configuration in a simple, declarative syntax. This configuration is then used to create machine images for multiple environments and cloud providers. The book begins by showing you how to create your first manifest while helping you understand the available components. You’ll then configure the most common built-in builder options for Packer and use runtime provisioners to reconfigure a source image for desired tasks. You’ll also learn how to control logging for troubleshooting errors in complex builds and explore monitoring options for multiple logs at once. As you advance, you’ll build on your initial manifest for a local application that’ll easily migrate to another builder or cloud. The chapters also help you get to grips with basic container image options in different formats while scaling large builds in production. Finally, you’ll develop a life cycle and retention policy for images, automate packer builds, and protect your production environment from nefarious plugins. By the end of this book, you’ll be equipped to smoothen collaboration and reduce the risk of errors by creating machine images consistently and automatically based on your defined configuration.
Table of Contents (18 chapters)
1
Part 1: Packer’s Beginnings
7
Part 2: Managing Large Environments
11
Part 3: Advanced Customized Packer

Preventing parallel processes from causing DoS

Recursively executing parallel builds is the potentially equivalent of a fork bomb for your Packer build. On the one hand, there are a lot more cores available in today’s machines. On the other hand, the cloud provides virtually unlimited threads and cores to an image builder. Each Packer template may itself contain many build sources. Parallelism and recursion together can quickly leave resources overwhelmed and unresponsive. It becomes necessary to limit the number of processes active at any given time.

We will adapt the build script from the previous chapter to use parallelism for each infrastructure grouping. Before we can automate builds in the next chapter, we also need to prepare for another potential problem — overlapping build pipelines, which must not be allowed. If a fix or change is made to code during an active build, then that build must be stopped before we can re-attempt a new build. Luckily, the GNU parallel...