Hacker News new | past | comments | ask | show | jobs | submit login

From all of the low-ops K8s distributions, k3s[0] is the best from perspective of inital setup, maintenance and usage on less powerful hardware.

There are even now higher-level tools such as k3os and k3sup to further reduce the initial deployment pains.

MicroK8s prides with 'No APIs added or removed'. That's not that positive in my book. K3s on the other hand actively removes the alpha APIs to reduce the binary size and memory usage. Works great if you only use stable Kubernetes primitives.

[0] https://k3s.io/

Thanks for mentioning K3sup [0]

I used Microk8s on a client project late last year and it was really painful, but I am sure it serves a particular set of users who are very much into the Snap/Canonical ecosystem.

In contrast, K3s is very light-weight and can be run in a container via the K3d project.

If folks want to work with K8s upstream or development patches against Kubernetes, they will probably find that KinD is much quicker and easier.

Minikube has also got a lot of love recently, and can run without having a dependency on Virtual Box too.

[0] https://k3sup.dev/ [1] https://kind.sigs.k8s.io/docs/user/quick-start/

This looks interesting, what control plane do the nodes usually connect to? I'm trying to see the use case for me, where I have a main NAS in my house and a few disparate Raspberry Pis, but I'm not sure if I would run the control plane on the NAS or if I would use a hosted one somewhere else.

I've had a number of issues with k3s on very low spec hardware (typically ARM), where it would take up to 25-50% of CPU just sitting idle with no pods. Stopped using it for those scenarios a year ago, wonder if that's fixed.

I had the same issue, it wasn't fixed on my last upgrade. I just let it eat some cpu : my pi is somewhat busy anyway

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact