WIP: live migration to a named node

Follow-up and derived from:
kubevirt#10712
Implements:
kubevirt/community#320

TODO: add functional tests

Signed-off-by: zhonglin6666 <zhangzl06@chinatelecom.cn>
Signed-off-by: Simone Tiraboschi <stirabos@redhat.com>

api/openapi-spec/swagger.json

@@ -14986,6 +14986,10 @@
      "kind": {
       "description": "Kind is a string value representing the REST resource this object represents. Servers may infer this from the endpoint the client submits requests to. Cannot be updated. In CamelCase. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#types-kinds",
       "type": "string"
+     },
+     "nodeName": {
+      "description": "NodeName is a request to try to migrate this VMI to a specific node. If it is non-empty, the migration controller simply try to configure the target VMI pod to be started onto that node, assuming that it fits resource, limits and other node placement constraints; it will override nodeSelector and affinity and anti-affinity rules set on the VM. If it is empty, recommended, the scheduler becomes responsible for finding the best Node to migrate the VMI to.",
+      "type": "string"
      }
     }
    },
@@ -16717,6 +16721,10 @@
    "v1.VirtualMachineInstanceMigrationSpec": {
     "type": "object",
     "properties": {
+     "nodeName": {
+      "description": "NodeName is a request to try to migrate this VMI to a specific node. If it is non-empty, the migration controller simply try to configure the target VMI pod to be started onto that node, assuming that it fits resource, limits and other node placement constraints; it will override nodeSelector and affinity and anti-affinity rules set on the VM. If it is empty, recommended, the scheduler becomes responsible for finding the best Node to migrate the VMI to.",
+      "type": "string"
+     },
      "vmiName": {
       "description": "The name of the VMI to perform the migration on. VMI must exist in the migration objects namespace",
       "type": "string"

pkg/virt-api/rest/subresource.go

@@ -303,7 +303,8 @@ func (app *SubresourceAPIApp) MigrateVMRequestHandler(request *restful.Request,
 				GenerateName: "kubevirt-migrate-vm-",
 			},
 			Spec: v1.VirtualMachineInstanceMigrationSpec{
-				VMIName: name,
+				VMIName:  name,
+				NodeName: bodyStruct.NodeName,
 			},
 		}, k8smetav1.CreateOptions{DryRun: bodyStruct.DryRun})
 		if err != nil {

pkg/virt-api/webhooks/validating-webhook/admitters/migration-create-admitter.go

@@ -116,6 +116,18 @@ func (admitter *MigrationCreateAdmitter) Admit(ctx context.Context, ar *admissio
 		return webhookutils.ToAdmissionResponseError(err)
 	}
 
+	nodeName := migration.Spec.NodeName
+	if nodeName != "" {
+		_, err = admitter.VirtClient.CoreV1().Nodes().Get(context.Background(), nodeName, metav1.GetOptions{})
+		if err != nil {
+			return webhookutils.ToAdmissionResponseError(err)
+		}
+
+		if nodeName == vmi.Status.NodeName {
+			return webhookutils.ToAdmissionResponseError(fmt.Errorf("the same source and destination node, so there is no need to migrate"))
+		}
+	}
+
 	reviewResponse := admissionv1.AdmissionResponse{}
 	reviewResponse.Allowed = true
 	return &reviewResponse

pkg/virt-api/webhooks/validating-webhook/admitters/migration-create-admitter_test.go

@@ -22,6 +22,7 @@ package admitters
 import (
 	"context"
 	"encoding/json"
+	"fmt"
 
 	"github.com/golang/mock/gomock"
 	. "github.com/onsi/ginkgo/v2"
@@ -30,10 +31,11 @@ import (
 	k8sv1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
-
-	"kubevirt.io/client-go/api"
+	"k8s.io/client-go/kubernetes/fake"
+	"k8s.io/client-go/testing"
 
 	v1 "kubevirt.io/api/core/v1"
+	"kubevirt.io/client-go/api"
 	"kubevirt.io/client-go/kubecli"
 
 	"kubevirt.io/kubevirt/pkg/virt-api/webhooks"
@@ -46,6 +48,8 @@ var _ = Describe("Validating MigrationCreate Admitter", func() {
 	var migrationCreateAdmitter *MigrationCreateAdmitter
 	var migrationInterface *kubecli.MockVirtualMachineInstanceMigrationInterface
 	var mockVMIClient *kubecli.MockVirtualMachineInstanceInterface
+	//var nodeSource *framework.FakeControllerSource
+	//var nodeInformer cache.SharedIndexInformer
 
 	BeforeEach(func() {
 		ctrl = gomock.NewController(GinkgoT())
@@ -97,7 +101,6 @@ var _ = Describe("Validating MigrationCreate Admitter", func() {
 
 		BeforeEach(func() {
 			migrationInterface.EXPECT().List(gomock.Any(), gomock.Any()).Return(&v1.VirtualMachineInstanceMigrationList{}, nil).MaxTimes(1)
-
 		})
 
 		It("should reject invalid Migration spec on create", func() {
@@ -258,6 +261,153 @@ var _ = Describe("Validating MigrationCreate Admitter", func() {
 			Expect(resp.Result.Message).To(ContainSubstring("DisksNotLiveMigratable"))
 		})
 
+		It("accept Migration spec on create when destination node exists", func() {
+			vmiName := "testmigratevmi6"
+			nodeName := "destmigratenode6"
+			vmi := api.NewMinimalVMI(vmiName)
+			vmi.Status.Phase = v1.Running
+			vmi.Status.Conditions = []v1.VirtualMachineInstanceCondition{
+				{
+					Type:   v1.VirtualMachineInstanceReady,
+					Status: k8sv1.ConditionTrue,
+				},
+			}
+
+			mockVMIClient.EXPECT().Get(context.Background(), vmi.Name, gomock.Any()).Return(vmi, nil)
+
+			kubeClient := fake.NewSimpleClientset()
+			virtClient.EXPECT().CoreV1().Return(kubeClient.CoreV1()).AnyTimes()
+
+			kubeClient.Fake.PrependReactor("get", "nodes", func(action testing.Action) (handled bool, obj runtime.Object, err error) {
+				return true, nil, fmt.Errorf("not found")
+			})
+
+			migration := v1.VirtualMachineInstanceMigration{
+				ObjectMeta: metav1.ObjectMeta{
+					Namespace: "default",
+				},
+				Spec: v1.VirtualMachineInstanceMigrationSpec{
+					VMIName:  vmiName,
+					NodeName: nodeName,
+				},
+			}
+			migrationBytes, _ := json.Marshal(&migration)
+
+			ar := &admissionv1.AdmissionReview{
+				Request: &admissionv1.AdmissionRequest{
+					Resource: webhooks.MigrationGroupVersionResource,
+					Object: runtime.RawExtension{
+						Raw: migrationBytes,
+					},
+				},
+			}
+
+			resp := migrationCreateAdmitter.Admit(context.Background(), ar)
+			Expect(resp.Allowed).To(BeFalse())
+			Expect(resp.Result.Message).To(ContainSubstring("not found"))
+		})
+
+		It("should reject Migration spec on create when the source and destination are the same node", func() {
+			vmiName := "testmigratevmi7"
+			nodeName := "destmigratenode7"
+			vmi := api.NewMinimalVMI(vmiName)
+			vmi.Status.Phase = v1.Running
+			vmi.Status.Conditions = []v1.VirtualMachineInstanceCondition{
+				{
+					Type:   v1.VirtualMachineInstanceReady,
+					Status: k8sv1.ConditionTrue,
+				},
+			}
+			vmi.Status.NodeName = nodeName
+
+			mockVMIClient.EXPECT().Get(context.Background(), vmi.Name, gomock.Any()).Return(vmi, nil)
+
+			kubeClient := fake.NewSimpleClientset()
+			virtClient.EXPECT().CoreV1().Return(kubeClient.CoreV1()).AnyTimes()
+
+			kubeClient.Fake.PrependReactor("get", "nodes", func(action testing.Action) (handled bool, obj runtime.Object, err error) {
+				return true, &k8sv1.Node{
+					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeName,
+					},
+				}, nil
+			})
+
+			migration := v1.VirtualMachineInstanceMigration{
+				ObjectMeta: metav1.ObjectMeta{
+					Namespace: "default",
+				},
+				Spec: v1.VirtualMachineInstanceMigrationSpec{
+					VMIName:  vmiName,
+					NodeName: nodeName,
+				},
+			}
+			migrationBytes, _ := json.Marshal(&migration)
+
+			ar := &admissionv1.AdmissionReview{
+				Request: &admissionv1.AdmissionRequest{
+					Resource: webhooks.MigrationGroupVersionResource,
+					Object: runtime.RawExtension{
+						Raw: migrationBytes,
+					},
+				},
+			}
+
+			resp := migrationCreateAdmitter.Admit(context.Background(), ar)
+			Expect(resp.Allowed).To(BeFalse())
+			Expect(resp.Result.Message).To(ContainSubstring("the same source and destination node"))
+		})
+
+		It("should accept Migration spec on create when the source and destination are not the same node", func() {
+			vmiName := "testmigratevmi7"
+			nodeName := "destmigratenode7"
+			vmi := api.NewMinimalVMI(vmiName)
+			vmi.Status.Phase = v1.Running
+			vmi.Status.Conditions = []v1.VirtualMachineInstanceCondition{
+				{
+					Type:   v1.VirtualMachineInstanceReady,
+					Status: k8sv1.ConditionTrue,
+				},
+			}
+			vmi.Status.NodeName = "sourcemigratenode"
+
+			mockVMIClient.EXPECT().Get(context.Background(), vmi.Name, gomock.Any()).Return(vmi, nil)
+
+			kubeClient := fake.NewSimpleClientset()
+			virtClient.EXPECT().CoreV1().Return(kubeClient.CoreV1()).AnyTimes()
+
+			kubeClient.Fake.PrependReactor("get", "nodes", func(action testing.Action) (handled bool, obj runtime.Object, err error) {
+				return true, &k8sv1.Node{
+					ObjectMeta: metav1.ObjectMeta{
+						Name: nodeName,
+					},
+				}, nil
+			})
+
+			migration := v1.VirtualMachineInstanceMigration{
+				ObjectMeta: metav1.ObjectMeta{
+					Namespace: "default",
+				},
+				Spec: v1.VirtualMachineInstanceMigrationSpec{
+					VMIName:  vmiName,
+					NodeName: nodeName,
+				},
+			}
+			migrationBytes, _ := json.Marshal(&migration)
+
+			ar := &admissionv1.AdmissionReview{
+				Request: &admissionv1.AdmissionRequest{
+					Resource: webhooks.MigrationGroupVersionResource,
+					Object: runtime.RawExtension{
+						Raw: migrationBytes,
+					},
+				},
+			}
+
+			resp := migrationCreateAdmitter.Admit(context.Background(), ar)
+			Expect(resp.Allowed).To(BeTrue())
+		})
+
 		DescribeTable("should reject documents containing unknown or missing fields for", func(data string, validationResult string, gvr metav1.GroupVersionResource, review func(ctx context.Context, ar *admissionv1.AdmissionReview) *admissionv1.AdmissionResponse) {
 			input := map[string]interface{}{}
 			json.Unmarshal([]byte(data), &input)

pkg/virt-controller/watch/migration.go

@@ -678,6 +678,7 @@ func (c *MigrationController) createTargetPod(migration *virtv1.VirtualMachineIn
 
 	templatePod.ObjectMeta.Labels[virtv1.MigrationJobLabel] = string(migration.UID)
 	templatePod.ObjectMeta.Annotations[virtv1.MigrationJobNameAnnotation] = migration.Name
+	templatePod.Spec.NodeName = migration.Spec.NodeName
 
 	// If cpu model is "host model" allow migration only to nodes that supports this cpu model
 	if cpu := vmi.Spec.Domain.CPU; cpu != nil && cpu.Model == virtv1.CPUModeHostModel {

pkg/virt-operator/resource/generate/components/crds.go

@@ -330,6 +330,8 @@ func NewVirtualMachineInstanceMigrationCrd() (*extv1.CustomResourceDefinition, e
 				Description: "The current phase of VM instance migration"},
 			{Name: "VMI", Type: "string", JSONPath: ".spec.vmiName",
 				Description: "The name of the VMI to perform the migration on"},
+			{Name: "NodeName", Type: "string", JSONPath: ".spec.nodeName",
+				Description: "The name of the the destination node to perform the migration on"},
 		}, &extv1.CustomResourceSubresources{
 			Status: &extv1.CustomResourceSubresourceStatus{},
 		})

pkg/virt-operator/resource/generate/components/validations_generated.go

@@ -13774,6 +13774,14 @@ var CRDsValidation map[string]string = map[string]string{
       type: object
     spec:
       properties:
+        nodeName:
+          description: |-
+            NodeName is a request to try to migrate this VMI to a specific node.
+            If it is non-empty, the migration controller simply try to configure the target VMI pod to be started onto that node,
+            assuming that it fits resource, limits and other node placement constraints; it will override nodeSelector and affinity
+            and anti-affinity rules set on the VM.
+            If it is empty, recommended, the scheduler becomes responsible for finding the best Node to migrate the VMI to.
+          type: string
         vmiName:
           description: The name of the VMI to perform the migration on. VMI must exist
             in the migration objects namespace

pkg/virtctl/vm/migrate.go

@@ -32,7 +32,10 @@ import (
 
 const COMMAND_MIGRATE = "migrate"
 
+var nodeName string
+
 func NewMigrateCommand(clientConfig clientcmd.ClientConfig) *cobra.Command {
+
 	cmd := &cobra.Command{
 		Use:     "migrate (VM)",
 		Short:   "Migrate a virtual machine.",
@@ -43,6 +46,8 @@ func NewMigrateCommand(clientConfig clientcmd.ClientConfig) *cobra.Command {
 			return c.migrateRun(args)
 		},
 	}
+
+	cmd.Flags().StringVar(&nodeName, "nodeName", nodeName, "--nodeName=<nodeName>: Flag to migrate this VM to a specific node regardless of its affinity rules. If it's omitted, recommended, the scheduler becomes responsible for finding the best Node to migrate the VM to.")
 	cmd.Flags().BoolVar(&dryRun, dryRunArg, false, dryRunCommandUsage)
 	cmd.SetUsageTemplate(templates.UsageTemplate())
 	return cmd
@@ -58,7 +63,7 @@ func (o *Command) migrateRun(args []string) error {
 
 	dryRunOption := setDryRunOption(dryRun)
 
-	err = virtClient.VirtualMachine(namespace).Migrate(context.Background(), vmiName, &v1.MigrateOptions{DryRun: dryRunOption})
+	err = virtClient.VirtualMachine(namespace).Migrate(context.Background(), vmiName, &v1.MigrateOptions{DryRun: dryRunOption, NodeName: nodeName})
 	if err != nil {
 		return fmt.Errorf("Error migrating VirtualMachine %v", err)
 	}

staging/src/kubevirt.io/api/core/v1/types.go

@@ -1361,6 +1361,14 @@ type VirtualMachineInstanceMigrationList struct {
 type VirtualMachineInstanceMigrationSpec struct {
 	// The name of the VMI to perform the migration on. VMI must exist in the migration objects namespace
 	VMIName string `json:"vmiName,omitempty" valid:"required"`
+
+	// NodeName is a request to try to migrate this VMI to a specific node.
+	// If it is non-empty, the migration controller simply try to configure the target VMI pod to be started onto that node,
+	// assuming that it fits resource, limits and other node placement constraints; it will override nodeSelector and affinity
+	// and anti-affinity rules set on the VM.
+	// If it is empty, recommended, the scheduler becomes responsible for finding the best Node to migrate the VMI to.
+	// +optional
+	NodeName string `json:"nodeName,omitempty"`
 }
 
 // VirtualMachineInstanceMigrationPhaseTransitionTimestamp gives a timestamp in relation to when a phase is set on a vmi
@@ -2236,6 +2244,13 @@ type MigrateOptions struct {
 	// +optional
 	// +listType=atomic
 	DryRun []string `json:"dryRun,omitempty" protobuf:"bytes,1,rep,name=dryRun"`
+	// NodeName is a request to try to migrate this VMI to a specific node.
+	// If it is non-empty, the migration controller simply try to configure the target VMI pod to be started onto that node,
+	// assuming that it fits resource, limits and other node placement constraints; it will override nodeSelector and affinity
+	// and anti-affinity rules set on the VM.
+	// If it is empty, recommended, the scheduler becomes responsible for finding the best Node to migrate the VMI to.
+	// +optional
+	NodeName string `json:"nodeName,omitempty"`
 }
 
 // VirtualMachineInstanceGuestAgentInfo represents information from the installed guest agent