Gremoire

How to Create Kubernetes Nodes and Schedule Pods Using Proxmox and Ubuntu VMs ​

This guide will walk you through how to:

  • Create multiple Ubuntu VMs in Proxmox
  • Use each VM as a Kubernetes node
  • Deploy pods across your cluster
  • Understand how resources affect pod scheduling

1. Setup Overview ​

What You’ll Build: ​

text
[ Proxmox Host ]
  ├─ VM 1: k8s-master-1 (Control Plane Node)
  ├─ VM 2: k8s-worker-1 (Worker Node)
  ├─ VM 3: k8s-worker-2 (Worker Node)

Each VM will run Ubuntu 22.04 with RKE2 or kubeadm installed.

2. Create VMs in Proxmox (Kubernetes Nodes) ​

VM Specs per node: ​

  • CPU: 2–4 vCPUs
  • RAM: 4–8 GB
  • Disk: 20–40 GB
  • Network: Bridged (e.g., vmbr0)

Create 3 VMs in the Proxmox UI:

  • k8s-master-1
  • k8s-worker-1
  • k8s-worker-2

Install Ubuntu Server 22.04 in each VM, and set a static IP.

3. Install RKE2 on All VMs ​

On Master Node: ​

bash
curl -sfL https://get.rke2.io | sudo sh -
sudo systemctl enable rke2-server.service --now

Get the token:

bash
sudo cat /var/lib/rancher/rke2/server/node-token

On Worker Nodes: ​

Create a file /etc/rancher/rke2/config.yaml:

yaml
server: https://<master-ip>:9345
token: <paste-token>

Install RKE2 Agent:

bash
curl -sfL https://get.rke2.io | sudo sh -
sudo systemctl enable rke2-agent.service --now

4. Access Kubernetes Cluster ​

On the master node:

bash
export KUBECONFIG=/etc/rancher/rke2/rke2.yaml
kubectl get nodes

You should see:

NAME            STATUS   ROLES           AGE   VERSION
k8s-master-1    Ready    control-plane   5m    v1.28.x
k8s-worker-1    Ready    <none>          3m    v1.28.x
k8s-worker-2    Ready    <none>          2m    v1.28.x

5. Deploy Pods and Observe Scheduling ​

Example Deployment: ​

bash
kubectl create deployment hello-world --image=nginx
kubectl scale deployment hello-world --replicas=10

Check pod distribution: ​

bash
kubectl get pods -o wide

You'll see pods spread across all worker nodes. Kubernetes schedules them based on available CPU and RAM.

6. Pod Scheduling Logic ​

Pods get scheduled based on:

  • Resource requests and limits
  • Available CPU/RAM on each node
  • Node taints/labels or affinity/anti-affinity rules

Example: If a pod requests 500m CPU and 512Mi RAM, a node with 2 CPUs and 2GB RAM can run around 4 such pods.

Summary ​

  • A Proxmox VM = 1 Kubernetes Node
  • More nodes → more pods can be scheduled
  • Use kubectl get nodes, top, describe to monitor usage
  • You control scaling by adding nodes or adjusting resources

Would you like an automated script to spin up VMs and join them to the cluster?