I use a single node microk8s instance mainly so I can stay familiar with syntax and things for work rather than an actual high availability system.

But I use Portainer and store my files on Github so Portainer can auto-update a deployment if I submit a code change, so kind of rudimentary CI/CD which could be fleshed out more with some GitHub Actions probably.

I use iSCSI mounted storage from my NAS on the host and k8s volumes storing configs there. Actual app data is on the NAS accessed via NFS from the relevant apps.

So a new deployment is usually test locally on my laptop, once it's good commit the code to github and either let the deployment auto update or go to Portainer and do it manually if it's a new deployment. Ingress traffic is done via Cloudflare Tunnels deployed in k8s.

I keep most apps in a single namespace called prod unless they need more than 1-2 pods. If I was doing this again I'd use a namespace per app, I do use a dedicated namespace for anything with a Helm deployment or needs a lot of pods (e.g. Immich)