PuppetConf 2017

How different is your development environment from production? I will share some techniques we can all use to model production service topologies on developer machines, so we can iterate on the same kinds of services that we use in production. I'll also demonstrate using labels and scheduling containers on mock-infrastructure and the interesting wins this technique buys us. It’s hard to have guarantees with anything in a deploy pipeline. We’ve all faced unexpected issues with drift between our iteration environments (dev/test/qa) and our deployment target for production services. Issues stemming from these differences can be super hard to track, and they can cause nasty bugs running through what is often some of the most untested code around: error-handling and recovery-automation. I will talk through how we can use containers on developer laptops to mock out our production infrastructure. When you mock your infrastructure and make it easy to visualize and query what’s happening, you open up a channel for empathy and enable people to do wonderful things! With these techniques you can: * Develop new functionality with more confidence it will function as intended * Model and Test against different topologies (ie: mid-deployment, mid-failure!) * Deploy using the same automation you use to iterate (dev/ci) plus it’s all repeatable. This is especially useful if you do not own your production environment, and is a surefire way to reduce pain and error. Environment differences cause pain. You want 0 drift between your environments. Modern tooling lets us model our production data centers on our personal machines, so we can iterate more quickly, experiment with failure modes, and deploy with confidence.