PuppetConf 2016

In 2014 KPN decided to use Puppet to manage thousands of its customers' servers. To guarantee a smooth transition to a puppetized infrastructure we needed to deliver high-quality modules. To that end we implemented a puppetized CI pipeline that included full acceptance tests on all target platforms (Linux AND Windows). In this session we'll take a closer look at the Jenkins, Beaker and VMpooler setup we use to provide this test environment. We'll talk about technical challenges and other lessons learned during this project.

I hope my audience will leave this talk with an understanding of how to monitor and tune performance for both Puppet and PE installations. Specific topics include: * JVM memory limits and how to manage them using Puppet. * How to monitor and tune Puppet Server to manage JRuby usage and profile catalog compilation. * How to monitor and tune PuppetDB to manage command queue depth and database disk usage. * How to tune the Enterprise console to manage the demands of classifier sync. * How to tune agents using splay and direct puppet such that Puppet infrastructure is used efficiently.

Yelp's PaaS based on Mesos is nice for deploying stateless application services, but we've found Puppet to be our tool of choice for managing clusters of heavily stateful applications like Zookeeper, Cassandra, and Elasticsearch, while supporting a high degree of configuration diversity. In this talk, I'll show how we have built "cluster native" Puppet modules for flexibility and scale. Our design allows us to: * Configure instances of a cluster independently, while still sharing core automation. * Dynamically scale clusters up and down as needed within minutes. * Provide control and tooling over these clusters to developers themselves. Along the way I'll show how you can tie together advanced Puppet techniques like Module Hiera Data, ENC, and custom types/functions to create robust automation.

With Puppet, you can declare what you want your infrastructure to look like and Puppet will take action, making the necessary changes to bring it to fruition. How do you know if your Puppet declarations have produced a working stack? Will future changes impact the health of your stack? In this presentation, I will discuss common pitfalls and failure cases when using Puppet to manage infrastructure and services, and how they can be mitigated or avoided with monitoring. I will talk about Sensu, a monitoring tool designed from the ground up to work in conjunction with Puppet. I will explain how Puppet and Sensu interact with one another to help produce reliable services. This presentation will include examples of Puppet & Sensu configuration and code.

There are few device-based modules in the wild, and there is little documentation for them. This talk will give insight into device-based modules, using a simple framework to quickly build a device-based module (agentless Puppet). To give quick results and getting something immediately useful, it uses the internet-enabled Philips Hue light bulb as the central use case. It shows how simple and how little code is required to interact with the philips hub using http / rest / json. This should show how easy it is to start once you have a code skeleton, but how easy it is for others to add functionality once the skeleton of the code is there.

Have you, a Puppet user, ever wondered how Puppet ships software? My goal here is to be fearlessly transparent and share the way we apply practices from the scientific method, and continuous delivery to deliver Puppet and friends on 80+ platform targets. This is a walk-through in DevOps and software delivery workflows. I'll trace the process of building out a new platform within the ecosystem which includes lifecycle selection, toolchain management, test tiering, and work scheduling. I'll take some planned detours into some of our freely available tools, including Vanagon and VMpooler. Weaved throughout the presentation are the roles people and teams play, how each person contributes to getting the bits out on time, and how our processes have evolved over the last five years.

At last year's conference, Puppet introduced new capabilities including a shift in how Puppet makes change to infrastructure from eventual convergence to an on-demand mode that enforces change quickly, provides immediate feedback while doing it, and respects relationships defined between nodes using the new application management language features. In this session, we'll review these new capabilities (also check out Nick Lewis's talk on Direct Change Control) and discuss the tooling we're introducing to make the most of Puppet's new capabilities. We'll cover extension points like the environment catalog, static compiler, cached catalogs, and more. If you’ve just started to use Puppet or are a long time Puppeteer kicking off agent runs with MCollective, you'll learn what's coming and how to take advantage of it.

Network automation presents a series of unique challenges for vendors and customers alike. In this session we will discuss the similarities and differences of network versus traditional systems administration. There will be real world examples of the complexity and scale of network configurations with an emphasis of separating the "how" from the "what". I also intend to demonstrate how Puppet is accelerating the depth of network automation by bridging the power of the Puppet language with YANG network data models.

Puppet Enterprise is more than just Puppet Open Source + Support. If you’re a Puppet Enterprise user today, or interested in amplifying your success with Puppet broadly across your organization, this is the session for you. We will explain what's new and what's coming up in Puppet Enterprise and also illustrate how its features work together to solve your problems.