feat(network): support hotplug/hotunplug of additional network interfaces

build/components/versions.yml

@@ -3,7 +3,7 @@ firmware:
   libvirt: v10.9.0
   edk2: stable202411
 core:
-  3p-kubevirt: v1.6.2-v12n.24
+  3p-kubevirt: feat/core/network-hotplug-support
   3p-containerized-data-importer: v1.60.3-v12n.18
   distribution: 2.8.3
 package:

docs/USER_GUIDE.md

@@ -2991,7 +2991,8 @@ If you specify the main network, it must be the first entry in the `.spec.networ
 Important considerations when working with additional network interfaces:
 
 - The order of listing networks in `.spec.networks` determines the order in which interfaces are connected inside the virtual machine.
-- Adding or removing additional networks takes effect only after the VM is rebooted.
+- Adding or removing an additional network (`Network` or `ClusterNetwork`) on a running VM is applied live without reboot. ACPI indexes of existing interfaces are preserved across add/remove cycles, so interface names in the guest OS stay stable.
+- Adding or removing the main network (`type: Main`) still requires a VM reboot, because it is tied to the pod's primary network interface and cannot be reconfigured on a running pod.
 - To preserve the order of network interfaces inside the guest operating system, it is recommended to add new networks to the end of the `.spec.networks` list (do not change the order of existing ones).
 - Network security policies (NetworkPolicy) do not apply to additional network interfaces.
 - Network parameters (IP addresses, gateways, DNS, etc.) for additional networks are configured manually from within the guest OS (for example, using Cloud-Init).

docs/USER_GUIDE.ru.md

@@ -3025,7 +3025,8 @@ EOF
 Особенности и важные моменты работы с дополнительными сетевыми интерфейсами:
 
 - порядок перечисления сетей в `.spec.networks` определяет порядок подключения интерфейсов внутри виртуальной машины;
-- добавление или удаление дополнительных сетей вступает в силу только после перезагрузки ВМ;
+- добавление или удаление дополнительной сети (`Network` или `ClusterNetwork`) на работающей ВМ применяется без перезагрузки. ACPI-индексы существующих интерфейсов сохраняются при добавлении/удалении, поэтому имена интерфейсов в гостевой ОС остаются стабильными;
+- добавление или удаление основной сети (`type: Main`) по-прежнему требует перезагрузки ВМ, так как она связана с основным сетевым интерфейсом пода и не может быть изменена на работающем поде;
 - чтобы сохранить порядок сетевых интерфейсов внутри гостевой операционной системы, рекомендуется добавлять новые сети в конец списка `.spec.networks` (не менять порядок уже существующих);
 - политики сетевой безопасности (NetworkPolicy) не применяются к дополнительным сетевым интерфейсам;
 - параметры сети (IP-адреса, шлюзы, DNS и т.д.) для дополнительных сетей настраиваются вручную изнутри гостевой ОС (например, с помощью Cloud-Init).

images/virtualization-artifact/pkg/common/network/spec.go

@@ -30,6 +30,14 @@ func CreateNetworkSpec(vm *v1alpha2.VirtualMachine, vmmacs []*v1alpha2.VirtualMa
 	macPool := NewMacAddressPool(vm, vmmacs)
 	var specs InterfaceSpecList
 
+	if len(vm.Spec.Networks) == 0 {
+		specs = append(specs, createMainInterfaceSpec(v1alpha2.NetworksSpec{
+			Type: v1alpha2.NetworksTypeMain,
+			ID:   ptr.To(ReservedMainID),
+		}))
+		return specs
+	}
+
 	for _, net := range vm.Spec.Networks {
 		if net.Type == v1alpha2.NetworksTypeMain {
 			specs = append(specs, createMainInterfaceSpec(net))

images/virtualization-artifact/pkg/common/network/types.go

@@ -37,13 +37,16 @@ func HasMainNetworkStatus(networks []v1alpha2.NetworksStatus) bool {
 }
 
 func HasMainNetworkSpec(networks []v1alpha2.NetworksSpec) bool {
-	for _, network := range networks {
-		if network.Type == v1alpha2.NetworksTypeMain {
-			return true
+	return GetMainNetworkSpec(networks) != nil
+}
+
+func GetMainNetworkSpec(networks []v1alpha2.NetworksSpec) *v1alpha2.NetworksSpec {
+	for i := range networks {
+		if networks[i].Type == v1alpha2.NetworksTypeMain {
+			return &networks[i]
 		}
 	}
-
-	return false
+	return nil
 }
 
 type InterfaceSpec struct {

images/virtualization-artifact/pkg/controller/kvbuilder/kvvm.go

@@ -744,6 +744,15 @@ func (b *KVVM) ClearNetworkInterfaces() {
 	b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces = nil
 }
 
+func (b *KVVM) SetNetworkInterfaceAbsent(name string) {
+	for i, iface := range b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces {
+		if iface.Name == name {
+			b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces[i].State = virtv1.InterfaceStateAbsent
+			return
+		}
+	}
+}
+
 func (b *KVVM) SetNetworkInterface(name, macAddress string, acpiIndex int) {
 	net := virtv1.Network{
 		Name: name,
@@ -770,15 +779,16 @@ func (b *KVVM) SetNetworkInterface(name, macAddress string, acpiIndex int) {
 		iface.MacAddress = macAddress
 	}
 
-	ifaceExists := false
-	for _, i := range b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces {
-		if i.Name == name {
-			ifaceExists = true
+	updated := false
+	for i, existing := range b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces {
+		if existing.Name == name {
+			b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces[i] = iface
+			updated = true
 			break
 		}
 	}
 
-	if !ifaceExists {
+	if !updated {
 		b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces = append(b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces, iface)
 	}
 }

images/virtualization-artifact/pkg/controller/kvbuilder/kvvm_test.go

@@ -22,7 +22,9 @@ import (
 
 	corev1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/types"
+	virtv1 "kubevirt.io/api/core/v1"
 
+	"github.com/deckhouse/virtualization-controller/pkg/common/network"
 	"github.com/deckhouse/virtualization/api/core/v1alpha2"
 )
 
@@ -145,3 +147,94 @@ func TestSetOSType(t *testing.T) {
 		}
 	})
 }
+
+func newTestKVVM() *KVVM {
+	return NewEmptyKVVM(types.NamespacedName{Name: "test", Namespace: "default"}, KVVMOptions{
+		EnableParavirtualization: true,
+	})
+}
+
+func TestSetNetworkInterfaceAbsent(t *testing.T) {
+	b := newTestKVVM()
+	b.SetNetworkInterface("default", "", 1)
+	b.SetNetworkInterface("veth_n12345678", "aa:bb:cc:dd:ee:ff", 2)
+
+	b.SetNetworkInterfaceAbsent("veth_n12345678")
+
+	for _, iface := range b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces {
+		if iface.Name == "veth_n12345678" {
+			if iface.State != virtv1.InterfaceStateAbsent {
+				t.Errorf("expected State %q, got %q", virtv1.InterfaceStateAbsent, iface.State)
+			}
+			return
+		}
+	}
+	t.Error("interface veth_n12345678 not found")
+}
+
+func TestSetNetworkInterfaceReplacesExisting(t *testing.T) {
+	b := newTestKVVM()
+	b.SetNetworkInterface("veth_n12345678", "aa:bb:cc:dd:ee:ff", 2)
+	b.SetNetworkInterfaceAbsent("veth_n12345678")
+
+	b.SetNetworkInterface("veth_n12345678", "aa:bb:cc:dd:ee:ff", 2)
+
+	for _, iface := range b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces {
+		if iface.Name == "veth_n12345678" {
+			if iface.State != "" {
+				t.Errorf("expected empty State after re-add, got %q", iface.State)
+			}
+			return
+		}
+	}
+	t.Error("interface veth_n12345678 not found")
+}
+
+func TestSetNetworkMarksRemovedAsAbsent(t *testing.T) {
+	b := newTestKVVM()
+	b.SetNetworkInterface("default", "", 1)
+	b.SetNetworkInterface("veth_n12345678", "aa:bb:cc:dd:ee:ff", 2)
+
+	setNetwork(b, network.InterfaceSpecList{
+		{InterfaceName: "default", MAC: "", ID: 1},
+	})
+
+	found := false
+	for _, iface := range b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces {
+		if iface.Name == "veth_n12345678" {
+			found = true
+			if iface.State != virtv1.InterfaceStateAbsent {
+				t.Errorf("removed interface should have State %q, got %q", virtv1.InterfaceStateAbsent, iface.State)
+			}
+		}
+		if iface.Name == "default" && iface.State != "" {
+			t.Errorf("kept interface should have empty State, got %q", iface.State)
+		}
+	}
+	if !found {
+		t.Error("removed interface should be retained with absent state, not deleted")
+	}
+}
+
+func TestSetNetworkAddsNewInterface(t *testing.T) {
+	b := newTestKVVM()
+	b.SetNetworkInterface("default", "", 1)
+
+	setNetwork(b, network.InterfaceSpecList{
+		{InterfaceName: "default", MAC: "", ID: 1},
+		{InterfaceName: "veth_n12345678", MAC: "aa:bb:cc:dd:ee:ff", ID: 2},
+	})
+
+	found := false
+	for _, iface := range b.Resource.Spec.Template.Spec.Domain.Devices.Interfaces {
+		if iface.Name == "veth_n12345678" {
+			found = true
+			if iface.ACPIIndex != 2 {
+				t.Errorf("expected ACPIIndex 2, got %d", iface.ACPIIndex)
+			}
+		}
+	}
+	if !found {
+		t.Error("new interface should be added")
+	}
+}

images/virtualization-artifact/pkg/controller/kvbuilder/kvvm_utils.go

@@ -355,7 +355,17 @@ func ApplyMigrationVolumes(kvvm *KVVM, vm *v1alpha2.VirtualMachine, vdsByName ma
 }
 
 func setNetwork(kvvm *KVVM, networkSpec network.InterfaceSpecList) {
-	kvvm.ClearNetworkInterfaces()
+	desiredByName := make(map[string]struct{}, len(networkSpec))
+	for _, n := range networkSpec {
+		desiredByName[n.InterfaceName] = struct{}{}
+	}
+
+	for _, iface := range kvvm.Resource.Spec.Template.Spec.Domain.Devices.Interfaces {
+		if _, wanted := desiredByName[iface.Name]; !wanted {
+			kvvm.SetNetworkInterfaceAbsent(iface.Name)
+		}
+	}
+
 	for _, n := range networkSpec {
 		kvvm.SetNetworkInterface(n.InterfaceName, n.MAC, n.ID)
 	}

images/virtualization-artifact/pkg/controller/vm/internal/sync_kvvm.go

@@ -690,6 +690,23 @@ func (h *SyncKvvmHandler) applyVMChangesToKVVM(ctx context.Context, s state.Virt
 		h.recorder.Event(current, corev1.EventTypeNormal, v1alpha2.ReasonVMChangesApplied, message)
 		log.Debug(message, "vm.name", current.GetName(), "changes", changes)
 
+		if hasNetworkChange(changes) {
+			if err := h.patchPodNetworkAnnotation(ctx, s); err != nil {
+				return fmt.Errorf("unable to patch pod network annotation: %w", err)
+			}
+
+			ready, err := h.isNetworkReadyOnPod(ctx, s)
+			if err != nil {
+				return fmt.Errorf("unable to check pod network status: %w", err)
+			}
+			if !ready {
+				msg := "Waiting for SDN to configure network interfaces on the pod"
+				log.Info(msg)
+				h.recorder.Event(current, corev1.EventTypeNormal, v1alpha2.ReasonVMChangesApplied, msg)
+				return nil
+			}
+		}
+
 		if err := h.updateKVVM(ctx, s); err != nil {
 			return fmt.Errorf("unable to update KVVM using new VM spec: %w", err)
 		}
@@ -749,6 +766,69 @@ func (h *SyncKvvmHandler) isVMUnschedulable(
 	return false
 }
 
+func hasNetworkChange(changes vmchange.SpecChanges) bool {
+	for _, c := range changes.GetAll() {
+		if c.Path == "networks" {
+			return true
+		}
+	}
+	return false
+}
+
+func (h *SyncKvvmHandler) isNetworkReadyOnPod(ctx context.Context, s state.VirtualMachineState) (bool, error) {
+	pods, err := s.Pods(ctx)
+	if err != nil {
+		return false, err
+	}
+	if pods == nil || len(pods.Items) == 0 {
+		return false, nil
+	}
+	errMsg, err := extractNetworkStatusFromPods(pods)
+	if err != nil {
+		return false, err
+	}
+	return errMsg == "", nil
+}
+
+func (h *SyncKvvmHandler) patchPodNetworkAnnotation(ctx context.Context, s state.VirtualMachineState) error {
+	log := logger.FromContext(ctx)
+
+	pod, err := s.Pod(ctx)
+	if err != nil {
+		return err
+	}
+	if pod == nil {
+		return nil
+	}
+
+	current := s.VirtualMachine().Current()
+	vmmacs, err := s.VirtualMachineMACAddresses(ctx)
+	if err != nil {
+		return err
+	}
+
+	networkConfigStr, err := network.CreateNetworkSpec(current, vmmacs).ToString()
+	if err != nil {
+		return fmt.Errorf("failed to serialize network spec: %w", err)
+	}
+
+	if pod.Annotations[annotations.AnnNetworksSpec] == networkConfigStr {
+		return nil
+	}
+
+	patch := client.MergeFrom(pod.DeepCopy())
+	if pod.Annotations == nil {
+		pod.Annotations = make(map[string]string)
+	}
+	pod.Annotations[annotations.AnnNetworksSpec] = networkConfigStr
+	if err := h.client.Patch(ctx, pod, patch); err != nil {
+		return fmt.Errorf("failed to patch pod %s network annotation: %w", pod.Name, err)
+	}
+	log.Info("Patched pod network annotation", "pod", pod.Name, "networks", networkConfigStr)
+
+	return nil
+}
+
 // isPlacementPolicyChanged returns true if any of the Affinity, NodePlacement, or Toleration rules have changed.
 func (h *SyncKvvmHandler) isPlacementPolicyChanged(allChanges vmchange.SpecChanges) bool {
 	for _, c := range allChanges.GetAll() {

images/virtualization-artifact/pkg/controller/vmchange/comparator_pod_placement.go

@@ -19,6 +19,7 @@ package vmchange
 import (
 	"reflect"
 
+	"github.com/deckhouse/virtualization-controller/pkg/common/network"
 	"github.com/deckhouse/virtualization/api/core/v1alpha2"
 )
 
@@ -89,10 +90,10 @@ func compareNetworks(current, desired *v1alpha2.VirtualMachineSpec) []FieldChang
 	desiredValue := NewValue(desired.Networks, desired.Networks == nil, false)
 
 	action := ActionRestart
-	// During upgrade from 1.6.0 to 1.7.0, network interface IDs are auto-populated for all existing VMs in the cluster.
-	// This allows avoiding a virtual machine restart during the version upgrade.
 	if isOnlyNetworkIDAutofillChange(current.Networks, desired.Networks) {
 		action = ActionNone
+	} else if isOnlyNonMainNetworksChanged(current.Networks, desired.Networks) {
+		action = ActionApplyImmediate
 	}
 
 	return compareValues(
@@ -104,6 +105,23 @@ func compareNetworks(current, desired *v1alpha2.VirtualMachineSpec) []FieldChang
 	)
 }
 
+// isOnlyNonMainNetworksChanged returns true when the Main network is unchanged
+// between current and desired (so only non-Main networks differ).
+// Empty networks list is equivalent to having an implicit default Main.
+func isOnlyNonMainNetworksChanged(current, desired []v1alpha2.NetworksSpec) bool {
+	currentMain := network.GetMainNetworkSpec(current)
+	desiredMain := network.GetMainNetworkSpec(desired)
+	currentHasMain := currentMain != nil || len(current) == 0
+	desiredHasMain := desiredMain != nil || len(desired) == 0
+	if !currentHasMain || !desiredHasMain {
+		return false
+	}
+	if currentMain == nil || desiredMain == nil {
+		return true
+	}
+	return reflect.DeepEqual(*currentMain, *desiredMain)
+}
+
 func isOnlyNetworkIDAutofillChange(current, desired []v1alpha2.NetworksSpec) bool {
 	if len(current) != len(desired) {
 		return false