Occasionally, Kubernetes workloads require specialized nodes. Sometimes it’s for machine learning, or to save money through burstable node types, or maybe just to lock certain Pods to a dedicated subset of nodes. Thankfully, Kubernetes offers a few useful mechanisms to inform the scheduler how we’d like our workloads distributed: node-based or pod-based affinity rules along with taints and tolerations. I’ll go over how to use these briefly, but I use these frequently at work for numerous reasons. Recently, I realized something interesting about how Persistent Volume Claims (PVCs) work with dynamically provisioned storage like EBS volumes (meaning volumes that are created automatically by Kubernetes based on a StorageClass, rather than referencing existing volumes). The default behavior of a StorageClass is to immediately create a volume as soon as the PVC is created. This can have some consequences when trying to guide how Pods are scheduled.
Recently, I wrote an AWS Lambda function at work in Ruby but I didn’t have a handy tool for creating a project skeleton like
bundle gem does. That means nothing bootstrapped my testing for me. While copy+pasting code into
pry proved that my simple function worked, that wasn’t quite good enough. What I really wanted was the ability to use RSpec with my Lambda code. After a cursory search of the Internet for some examples, I was left disappointed with how little I found. So, I rolled up my sleeves and figured it out.
Amazon’s CloudFormation is a wonderful and flexible tool for provisioning and managing resources in an EC2 VPC. It really takes the concept of infrastructure-as-code and helps make it a reality. For all its flexibility though, it sacrifices intuitiveness and ease. It is also limited by the rigidity of JSON, which isn’t a full-fledged language so it doesn’t support variables (although Parameters, Mappings, and References to them are a long-winded and difficult to parse approach that comes close) or easily referencing reusable external libraries. It also isn’t possible to define arbitrary functions, iterate over lists, or define anything but the most rudimentary conditional sections. This is by no means a criticism of CloudFormation, as it has certainly done a lot to turn a serialization format into a pseudo scripting language, but these are my observations that might frustrate other people when using it.
That’s where ERB comes in. As a big advocate of Ruby, whenever I think of templating the first thing that comes to mind is ERB. A while back, I put together a super simple script that generates templates from JSON “layouts” and “snippets” (think views and partials from Rails), both of which fully support ERB and all of its Ruby goodness.
Continue reading Reusable CloudFormation Snippets with ERB