PoC: Add netplan as alternative backend to nmstate

AGENTS.md

@@ -183,6 +183,9 @@ Cluster configuration via environment variables:
 - `KUBECONFIG`: Path to kubeconfig (auto-detected via ./cluster/kubeconfig.sh)
 - `NMSTATE_VERSION`: When set to `latest`, uses nmstate-git from copr during `make cluster-up`
 - `NM_VERSION`: When set to `latest`, installs NetworkManager from copr networkmanager/NetworkManager-main repository during `make cluster-up`
+- `NETWORK_RENDERER`: Network renderer to use (default: `NetworkManager`, options: `NetworkManager`, `networkd`)
+  - `NetworkManager`: Uses NetworkManager for all network interfaces
+  - `networkd`: Uses systemd-networkd for eth1/eth2, NetworkManager for eth0 (cluster connectivity)
 
 Network interface names vary by provider:
 - k8s providers: eth0, eth1, eth2

BACKEND_SELECTION.md

@@ -0,0 +1,210 @@
+# Backend Selection for kubernetes-nmstate
+
+This document explains how to select the network configuration backend when deploying kubernetes-nmstate to a development cluster.
+
+## Available Backends
+
+- **nmstate** (default): Uses nmstatectl and NetworkManager for network configuration
+- **netplan**: Uses netplan's D-Bus API for network configuration (PoC)
+
+## Using the BACKEND Variable
+
+The `BACKEND` Makefile variable allows you to select which backend to use when deploying to a local development cluster.
+
+### Default Behavior (nmstate)
+
+```bash
+make cluster-sync
+```
+
+This deploys kubernetes-nmstate with the default nmstate backend.
+
+### Using Netplan Backend
+
+To deploy with the netplan backend, set the `BACKEND` variable:
+
+```bash
+BACKEND=netplan make cluster-sync
+```
+
+Or export it before running make:
+
+```bash
+export BACKEND=netplan
+make cluster-sync
+```
+
+## How It Works
+
+1. The `BACKEND` variable is defined in the Makefile (defaults to `nmstate`)
+2. It's exported so it's available to shell scripts
+3. The `cluster/sync.sh` script patches the NMState CR with the selected backend
+4. The operator configures handler pods with the appropriate `NMSTATE_BACKEND` environment variable
+5. Handlers initialize the selected backend at startup
+
+## Example Workflow
+
+### Testing with nmstate (default)
+
+```bash
+# Build images
+make handler operator
+
+# Deploy to cluster
+make cluster-sync
+
+# Verify deployment
+kubectl get nmstate nmstate -o jsonpath='{.spec.backend}'
+# Output: (empty - defaults to nmstate)
+
+# Check handler pods are using nmstate
+kubectl get pods -n nmstate -l component=kubernetes-nmstate-handler -o jsonpath='{.items[0].spec.containers[0].env[?(@.name=="NMSTATE_BACKEND")].value}'
+# Output: nmstate
+```
+
+### Testing with netplan
+
+```bash
+# Build images
+make handler operator
+
+# Deploy with netplan backend
+BACKEND=netplan make cluster-sync
+
+# Verify deployment
+kubectl get nmstate nmstate -o jsonpath='{.spec.backend}'
+# Output: netplan
+
+# Check handler pods are using netplan
+kubectl get pods -n nmstate -l component=kubernetes-nmstate-handler -o jsonpath='{.items[0].spec.containers[0].env[?(@.name=="NMSTATE_BACKEND")].value}'
+# Output: netplan
+
+# Monitor handler logs for netplan D-Bus interactions
+kubectl logs -n nmstate -l component=kubernetes-nmstate-handler -f
+```
+
+### Switching Backends
+
+To switch from one backend to another:
+
+```bash
+# Switch to netplan
+BACKEND=netplan make cluster-sync
+
+# The operator will update the NMState CR
+# Handler pods will be restarted with the new backend
+
+# Switch back to nmstate
+BACKEND=nmstate make cluster-sync
+# Or simply:
+make cluster-sync
+```
+
+## Implementation Details
+
+### Makefile
+
+```makefile
+# Backend selection for network configuration (nmstate or netplan)
+export BACKEND ?= nmstate
+```
+
+### cluster/sync.sh
+
+```bash
+function patch_handler_backend() {
+    if [ -n "${BACKEND}" ] && [ "${BACKEND}" != "nmstate" ]; then
+        echo "Patching NMState CR to use backend: ${BACKEND}"
+        $kubectl patch -f $nmstate_cr_manifest --patch "{\"spec\": {\"backend\": \"${BACKEND}\"}}" --type=merge
+    fi
+}
+```
+
+This function:
+- Only patches if `BACKEND` is set and not the default "nmstate"
+- Uses `kubectl patch` to update the NMState CR's `spec.backend` field
+- The operator watches for changes and updates handler pods accordingly
+
+## Requirements by Backend
+
+### nmstate Backend
+
+**Requirements:**
+- NetworkManager >= 1.22 running on nodes
+- nmstatectl CLI available
+- D-Bus system bus access
+
+**Already provided by default handler image:**
+- nmstate package
+- NetworkManager dependencies
+
+### netplan Backend (PoC)
+
+**Requirements:**
+- netplan >= 0.103 installed on nodes
+- netplan D-Bus service running (`io.netplan.Netplan`)
+- D-Bus system bus access (already available)
+- godbus/dbus/v5 Go library (needs to be added to go.mod)
+
+**Additional setup needed:**
+- Handler container image needs netplan package
+- Format conversion between nmstate and netplan YAML
+
+## Troubleshooting
+
+### Backend not applied
+
+If the backend doesn't change:
+
+```bash
+# Check NMState CR
+kubectl get nmstate nmstate -o yaml
+
+# Check if operator is running
+kubectl get pods -n nmstate -l component=kubernetes-nmstate-operator
+
+# Check operator logs
+kubectl logs -n nmstate -l component=kubernetes-nmstate-operator
+```
+
+### Handler pods not updated
+
+```bash
+# Restart handler pods to pick up new configuration
+kubectl delete pods -n nmstate -l component=kubernetes-nmstate-handler
+
+# Watch pods restart
+kubectl get pods -n nmstate -w
+```
+
+### Netplan backend errors
+
+```bash
+# Check if netplan D-Bus service is running on nodes
+./cluster/ssh.sh node01 "systemctl status netplan-dbus || systemctl status netplan"
+
+# Check D-Bus service availability
+./cluster/ssh.sh node01 "busctl list | grep netplan"
+
+# Check handler logs for D-Bus errors
+kubectl logs -n nmstate -l component=kubernetes-nmstate-handler | grep -i dbus
+```
+
+## Production Considerations
+
+**Note:** The `BACKEND` Makefile variable is for **development and testing only**. In production:
+
+1. Set the backend in the NMState CR directly:
+   ```yaml
+   apiVersion: nmstate.io/v1
+   kind: NMState
+   metadata:
+     name: nmstate
+   spec:
+     backend: netplan  # or nmstate
+   ```
+
+2. Ensure nodes have the required backend software installed
+3. Test thoroughly before deploying to production
+4. Monitor handler logs for backend-specific errors
+5. Have a rollback plan to switch backends if needed