I flew out to KubeCon + CloudNativeCon North America 2018 in December to give a talk on Single Sign-On for Kubernetes.
This talk was inspired by three articles that I originally wrote for The New Stack.
The New Stack kindly invited me to join them for an episode of their Context podcast. Each week they discuss with popular topics with a contributor and get to have a deeper discussion about the contributors work.
This week I discussed my experience with building custom controllers for Kubernetes and what I look forward to at KubeCon + CloudNativeCon North America 2018.
Ensuring that configuration is up to date is difficult with any infrastructure; With Kubernetes and the lack of versioned ConfigMaps, this problem is amplified.
At Pusher, we have a number of applications that can’t dynamically reload their configuration. Over the last two years, this has caused pain for our engineers and made deploying new configuration hard.
In this post I explore our new project Wave and how we solved the problem of ensuring that our application’s running configuration matches the desired configuration.
After integrating Spot Instances with Pusher’s Kubernetes Infrastructure, I share the lessons I learned about making Kubernetes workloads tolerant to node failures and as available as possible.