NurlashKO's Blog🐈‍

I got some inspiration after reading Reinventing Kubernetes in 2025: a post-mortem of my “simple” stack (and you should too).

Here I would like to share my plans for the future of this project.

First, let’s recall how it was working so far.

• Hosted on GCP instance
• Services built and pushed to GCR
• “Provisioner” is a daemon process responsible for the latest version of each service running based on predefined configuration.
• Internal networking fully relies on Docker Mesh while External traffic is routed through Nginx

Moving Forward: Transition to Kubernetes

• Hosted on a DigitalOcean Droplet (2 CPU, 2 GB RAM)
• Images pushed to RepoFlow
  • 10 GB of free storage!
• K8S deployed via K3S.

In essence, I’m replacing a bunch of Bash scripts with YAML configs. To my surprise, setting up the VM went incredibly smoothly.

Quite literally:

1. Spin up a new VM with open 6443 port
2. SSH & Run curl -sfL https://get.k3s.io | sh -
3. Copy contents of /etc/rancher/k3s/k3s.yaml to your local machine

Voilà, it simply works! 🤯

k create ns hello
k create deployment -n hello hello-world-dep --image=ollyw123/helloworld

I’ve already deployed all core components — blog, database, and vault. This setup allows me to write, save, and publish this very post.

Networking, certificate rotation, and volume provisioning work almost entirely out of the box.

Of course, there’s a trade-off: I’m gaining simplicity but losing transparency. I don’t fully understand how traefik forwards traffic, nor the details of how local-path-provisioner is implemented — but for now, as long as it works, that’s good enough 🙌

When I finally reached peace, I realized: It was never about escaping Kubernetes. It was about accepting that every abstraction eventually becomes it.