Configuration

Customize your k8s-provisioner cluster

k8s-provisioner uses two configuration files:

config.yaml

Main configuration file for Kubernetes and components.

cluster:
  name: "k8s-lab"
  pod_cidr: "10.244.0.0/16"
  service_cidr: "10.96.0.0/12"

versions:
  kubernetes: "1.32"
  crio: "v1.32"
  calico: "3.28.0"
  metallb: "0.14.8"
  istio: "1.28.2"
  karpor: "0.7.6"

network:
  interface: "eth1"
  controlplane_ip: "192.168.56.10"
  metallb_range: "192.168.56.200-192.168.56.250"

storage:
  nfs_server: "storage"
  nfs_path: "/exports/k8s-volumes"
  default_dynamic: true       # nfs-dynamic as default StorageClass

nodes:
  - name: "storage"
    role: "storage"
  - name: "controlplane"
    role: "controlplane"
  - name: "node01"
    role: "worker"
  - name: "node02"
    role: "worker"

components:
  cni: "calico"
  load_balancer: "metallb"
  service_mesh: "istio"
  monitoring: "prometheus-stack"  # Options: prometheus-stack, none
  logging: "loki"                 # Options: loki, none
  karpor: "enabled"               # Options: enabled, none

# Karpor AI configuration (optional)
karpor_ai:
  enabled: true
  backend: "ollama"           # Options: openai, azureopenai, huggingface, ollama
  model: "llama3.2:3b"        # Local model (or minimax-m2.5:cloud for cloud)
  auth_token: ""              # API token (not needed for ollama local)
  base_url: ""                # Custom endpoint (leave empty for default)

# Ollama cloud API key (only for :cloud models)
ollama:
  api_key: ""                 # Get from https://ollama.com/settings/keys

Configuration Options

components.karpor

Value	Description
`enabled`	Install Karpor with AI features
`none`	Skip Karpor installation

karpor_ai.enabled

Value	Description
`true`	Enable AI features with Ollama
`false`	Disable AI (Ollama not installed)

karpor_ai.model

Local models (run inside cluster):

Model	RAM Required	Quality
`llama3.2:1b`	~2GB	Basic
`llama3.2:3b`	~4GB	Good (default)
`qwen2.5-coder:7b`	~8GB	Excellent
`llama3.1:8b`	~10GB	Excellent

Cloud models (require API key):

Model	Description
`minimax-m2.5:cloud`	Top performer
`qwen3-coder:480b-cloud`	Excellent for code
`glm-4.7:cloud`	Good general purpose

storage.default_dynamic

Value	Description
`true`	`nfs-dynamic` is the default StorageClass
`false`	No default StorageClass

vagrant/settings.yaml

VM configuration for VirtualBox.

box_name: "bento/debian-12"
vm:
- name: "storage"
  ip: "192.168.56.20"
  memory: "1024"
  cpus: "1"
  role: "storage"
- name: "controlplane"
  ip: "192.168.56.10"
  memory: "6144"    # Extra for monitoring stack
  cpus: "4"
  role: "controlplane"
- name: "node01"
  ip: "192.168.56.11"
  memory: "8192"    # Extra for AI workloads (Ollama)
  cpus: "2"
  role: "worker"
- name: "node02"
  ip: "192.168.56.12"
  memory: "4096"
  cpus: "2"
  role: "worker"

VM Resource Allocation

VM	Memory	CPUs	Purpose
Storage	1 GB	1	NFS Server
ControlPlane	6 GB	4	K8s Master + Monitoring
Node01	8 GB	2	Worker + AI Workloads
Node02	4 GB	2	Worker
Total	19 GB	9

Note: Node01 has extra memory for Ollama AI workloads. If not using Karpor AI, you can reduce it to 4GB.

Minimal Configuration (Without Karpor)

If you have limited resources, disable Karpor and AI:

components:
  karpor: "none"

karpor_ai:
  enabled: false

And reduce VM resources in vagrant/settings.yaml:

- name: "controlplane"
  memory: "4096"
  cpus: "2"
- name: "node01"
  memory: "4096"
  cpus: "2"