Testing in production aka user-based dynamic api routing

We've been discussing testing in production for a long time. Especially for our "main" api we knew we needed to do something in order to improve the way we develop and deploy. This api relies on production (or production-like) data, and is also the primary contact point for a host of different clients (both apps, set-top boxes and web apps).

Migrating to Traefik 2.0 without downtime

We've been using Traefik as the Ingress Controller of choice ever since we started using Kubernetes. We also use Traefik for our non-containerized apps, where Consul is used as the "source of truth" for routing configuration.

Redesigning RiksTV's Kubernetes setup

Weird how time flies - we've actually been running Kubernetes in production for over 2 years already. Until now, we've used an Ansible-based provisioning approach written by yours truly (<https://github.com/trondhindenes/Kubelini>). This has served us well enough, but it has also required some maintenance effort so we were looking to simplify and standardize. This post is about that.