I've been getting in-depth with Kubernetes lately; from watching YouTube videos on real-life projects with it, to building a multi-node homelab with a K3s cluster. (And now another cluster using kubeadm)
Oracle VirtualBox to run all the virtual machines for the Kubernetes cluster
Most recently, I started building clusters from scratch using kubeadm on 4 Virtual Machines (1 control plane, 3 workers), and this is what I learned so far:
Ansible is your friend!
Ansible allows for easy configuration of all VMs with one command.
Along with Vagrant for easy VM provisioning, Ansible helps so much with running commands in parallel to make sure all the VMs are identically-configured 👌 Sure, you could run a pre-built image with kubeadm, kubectl, kubelet, and other configs pre-configured, but where's the fun in that? ;) Plus, configuring systems to be Kubernetes-ready is an extremely good skill to have for on-prem cluster administration!
Backing up and restoring etcd
The etcd backup process can be simplified with a CronJob and an external storage solution.
In managed environments like AKS (Azure), GKE (Google), or EKS (AWS), I didn't really focus on the inner workings of a cluster - I just wanted it to work. And that's actually one of the main benefits of using a managed service for K8s. But in a bare-metal environment, backing up etcd is essential. Think of it like the state file in Terraform - it stores information on what pods, deployments, services, etc are running in the cluster. Losing your etcd database means losing access to the core benefit of K8s: orchestration.
Linux helps. A ton!!
With Linux knowledge, you can use tools like sed to change properties in YAML manifests immediately.
Yes, knowing the kubeadm and kubectl commands makes your life easier. But do you know what makes this learning journey smoother? Knowing Linux! There's just so much file and directory manipulation, input redirections, piping, and Bash scripting that would improve your quality-of-life so much as a Kubernetes developer or administrator. Seriously, you can not skip straight to Kubernetes without first familiarizing yourself with Linux and the command line. It is night and day difference!
"Cattle, not pets"
Using Vagrant enables easy teardown (or bringing up) of your entire cluster or specific nodes.
I learned this saying, which is a design philosophy where you treat your systems that you build as cattle: easily replaceable and disposable, and not pets: precious, unique and irreplaceable. As soon as cattle (server, cluster, whatever) is lost, be ready to immediately replace it with a new identical (Or nearly identical) one. Practice GitOps and Infrastructure as Code to achieve this!
Conclusion
Excited to get even more in-depth, and go over more administrative concepts in K8s. Also looking forward to the security aspect of it!
