Service catalog seems to have quietly left the room. Gone but not forgotten, all aboard the operator hype train.
How to install Loki & Fluent Bit for lightweight monitoring across your k8s cluster.
cert-manager, one of the most important operators in the k8s ecosystem, just moved from Bazel to Make, and that's awesome. In case it wasn't clear, I'm biased in favor of Makefiles.
How I (painfully) upgraded a running k0s cluster from single-master to HA master. In the end I think a single well-equipped master was actually quite stable and better, but the lure of 'web scale' was too strong. Also, if you're thinking of attempting this, just start a new cluster.
Looking to run Postgres on ZFS? I've gathered some of the information and sage advice out there to give you a head start on figuring out how to do it safely and efficiently.
A while back, I had a somewhat soft node failure due to Calico not allowing the same node to come up under a different IP address.
I recently reworked my storage setup in pursuit of both production and flexibility. Looking at the F/OSS space led me to OpenEBS ZFS LocalPV with Longhorn on top, and here's why.
A quick and simple guide to automating your k0s cluster backup with ansible and systemd. You can probably guess the code you're about to read.
If you're running on Hetzner, you might want to update your DNS settings to use some other resolvers. This can problem can manifest if you're trying to resolve DNS names that resolve to your own machines *from inside your own machines*.
Want to wait for some Kubernetes resources and don't want to write a full reconciliation loop/operator? Here's how to hack it.
You may not need Flux/ArgoCD or any fancy automated reconciliation loop deployment process just yet! Plain old CI runners work quite well for simple automated deployments (this approach powers this blog right now!).
I finish up the benchmarking process, and share the results. If you want to find what I actually end up using, the answer is in here!
I describe in detail the YAML and Makefile scripts that power the fio and pgbench based tests that will run on every storage provider I managed to set up
I install even more storage providers -- this time OpenEBS cStor, Jiva, LocalPV hostPath, LocalPV ZFS and LINSTOR
I start installing storage plugins, this time Rook (Ceph) -- which takes a while -- and OpenEBS Mayastor. Read this if you want to read a feel-good underdog system administration story.
Part 1 of my second crack at benchmarking storage providers that scale from hobbyist to enterprise. In this post I go over the setup (the code isn't public just yet) of the dedicated servers I'll be using
Kubernetes upgrade sunday -- A series of posts where I do some maintenance and upgrades on my k8s cluster that sometimes happens on a sunday. This sunday I work on node local DNS caching, kube-router, jaeger, cert-manager, traefik, and containerd
Service to service communication broke down due to old Kubeconfig being fed to my old version of kube-router.
I run a small experiment to figure out what the optimal setting is for pm2.
Improving my security posture with ease by following best practices from a blog post (https://www.simplecto.com/get-more-secure-websites-from-your-traefik-configuration) thanks to thanks to Traefik and Kubernetes. This post is also unintentionally an exploration of the amount of work it takes to switch from Traefik v1 to v2.