Bare metal installation for homelab / high CPU usage

Rancher
1

Hello there,

after trying to find the best solution for me (minikube, docker swarm, microk8s, kind) I still ended up in trying rancher. It has a lot of good features and I think, this could rly be the way to go even for companies. Atm I just hear how complicated it is to use kubernates and so on but in the last time it seams to get better and better (with the scripts of rancher). Sadly in my opinion the documentation of the installation in rancher is still “complicated” and hard to find… After a while I came across some guides and tried it out. So far so good. Im pretty happy about it but have 1 last question… How can I rly use it for an single node installation without this 10%-15% CPU usage even with a fresh installation?
First I wanted to use k3s but I ended up using the default rke2 one. After reading the documentation about rke2 I think this should be good for my use case and seams to be the path to go with rancher.

But back to my question… Something is running in the background causing this cpu load that sucks…

My config:
Proxmox → (VM) Ubuntu Server - 20.x (CPU:12 ; 16 GB RAM ; q35 ; ovmf)

sudo su
mkdir -p /etc/rancher/rke2
cd /etc/rancher/rke2
nano config.yaml

token: xyz
tls-san:

  • 192.168.1.42
  • rancherino.example.ch
  • rancher.example.ch

curl -sfL https://get.rancher.io | sh -
rancherd --help
systemctl enable rancherd-server.service
systemctl start rancherd-server.service
journalctl -eu rancherd-server -f

rancherd reset-admin

https://192.168.1.42:8443

Now… what is causing this load?
kube-apiserver and etcd
As far as I did read is this an HA stuf which I wouldnt need.
My goal is just to have a single node lightweight rancher installed on this ubuntu machine. What can I do to reach this goals?

2

Sadly seams like not much support is gona come for “non enterprise” people. Even if it seams to be more an kubernates stuff, still In my opinion this reflects the opinion of the most small to medium companies who dont want to switch to a more complex stuff like this with an overhead of kubernates without the need for the most features.

I think, I ll stay with docker swarm and still recommend to other to go this way.

3

There’s not much of anything to “support” here, the K8s stack is what it is and if you’re not happy with 1/8th of one core then you’re just not going to be able to use it.