Kind (laptop lab)

Run VirtFoundry on your machine with kind (Kubernetes in Docker). You do not need a VLAN, a managed switch, or MetalLB.

This is the path for trying the UI and a first VM. Public guest IPs on your Wi-Fi are optional and come later on this page.

What “public” means here

On a homelab with a switch, public is often a dedicated VLAN. On kind it is either off (VMs on the pod network + noVNC in the UI) or a second Docker network that only your laptop can reach. Same Helm keys either way — see Topologies — public underlay.

What you need

Tool	Notes
Docker	Running
kind	kind version
kubectl + Helm 3
/dev/kvm	Linux — nested VMs are fast. Docker Desktop (macOS/Windows) usually has no KVM; use software emulation (slower).

# Linux
test -e /dev/kvm && echo "KVM ok" || echo "No KVM — you will use emulation"

---

1. Create the cluster

Save as kind-cluster.yaml (or copy examples/kind-cluster.yaml from the chart repo).

kind: Cluster
apiVersion: kind.x-k8s.io/v1alpha4
name: virtfoundry
nodes:
  - role: control-plane
    extraMounts:
      - hostPath: /dev/kvm
        containerPath: /dev/kvm
    extraPortMappings:
      - containerPort: 30880
        hostPort: 8080
        protocol: TCP

No /dev/kvm? Delete the extraMounts block.

kind create cluster --config kind-cluster.yaml
kubectl cluster-info --context kind-virtfoundry

Kind ships a standard StorageClass (local-path). Volume snapshots will not work; use VM snapshots in the UI.

---

2. Platform components

The Helm chart does not install KubeVirt, Multus, or CDI. From a helm-charts clone:

export KUBECONFIG="${KUBECONFIG:-$HOME/.kube/config}"
./scripts/setup/multus.sh
./scripts/setup/kubevirt.sh
./scripts/setup/cdi.sh

Without KVM, turn on emulation after KubeVirt is Available:

kubectl -n kubevirt patch kubevirt kubevirt --type merge -p \
  '{"spec":{"configuration":{"developerConfiguration":{"useEmulation":true}}}}'

Wait until kubectl get kubevirt -n kubevirt shows Available.

---

3. Install VirtFoundry (public off)

Overlay values-kind.yaml: NodePort 30880 (mapped to host 8080), public.enabled: false.

curl -fsSL https://raw.githubusercontent.com/virtfoundry/helm-charts/main/charts/virtfoundry/values-kind.yaml \
  -o values-kind.yaml

helm repo add virtfoundry https://virtfoundry.github.io/helm-charts
helm repo update

helm install virtfoundry virtfoundry/virtfoundry \
  --version 0.5.0 \
  --namespace virtfoundry-system --create-namespace \
  -f values-kind.yaml \
  --set secrets.rootPassword='change-me' \
  --set secrets.jwtSecret='change-me-long-random'

Open http://127.0.0.1:8080 — user root, password change-me.

Deploy a container-disk VM (Templates → small offering → Console). The guest has a pod IP only; you reach it through noVNC, not from your LAN.

Isolated VPCs still work: bridge-keeper creates virtfoundry-br0 inside the kind node. That L2 never leaves Docker.

---

4. Optional — local “public” IPs (still no VLAN)

Use this when you want a guest address you can ping from the same laptop, without a switch.

Kubernetes inside kind uses eth0 on the kind Docker network. Never set that as public.bridge.uplink — bridge-keeper would steal the node IP and the cluster dies.

Add a second Docker network (this is the kind equivalent of a second NIC on a house LAN):

docker network create --driver bridge \
  --subnet 10.0.50.0/24 --gateway 10.0.50.1 \
  virtfoundry-pub

docker network connect virtfoundry-pub virtfoundry-control-plane

docker exec virtfoundry-control-plane ip -br addr
# eth0 = kind (kubelet)     ← do not touch
# eth1 = 10.0.50.x          ← public uplink

If the extra NIC is not eth1, use the name you see.

# values-kind-public.yaml — merge with values-kind.yaml
platform:
  networking:
    public:
      enabled: true
      cidr: 10.0.50.0/24
      gateway: 10.0.50.1          # Docker bridge on your laptop
      dns:
        - 10.0.50.1
      ipPool:
        start: 10.0.50.10
        end: 10.0.50.99
      bridge:
        name: vf-pub0
        uplink: eth1              # extra NIC, not eth0
        address: 10.0.50.2/24     # host side of vf-pub0; not in the VM pool

helm upgrade virtfoundry virtfoundry/virtfoundry \
  --version 0.5.0 \
  -n virtfoundry-system \
  -f values-kind.yaml \
  -f values-kind-public.yaml \
  --reuse-values

Attach the public network on the VM (UI → Public / deploy with public IP). From the laptop:

ping -c 2 10.0.50.10    # first address in the pool, if allocated

Your phone on Wi-Fi will not see 10.0.50.0/24 unless you add a route. That is expected: kind public is local to Docker, like a lab VLAN that only exists on this machine.

Do not enslave eth0

uplink: eth0 takes down kubelet inside the kind node. Recover with kind delete cluster --name virtfoundry and recreate.

---

Cleanup

kind delete cluster --name virtfoundry
docker network rm virtfoundry-pub   # if you created it

Next

Goal	Doc
First VM clicks	Quickstart
Helm keys	Configuration — public networking
VLAN vs house LAN on real nodes	Topologies
Why KubeVirt / Multus / CDI	Installation