VPA: add InPlaceOrRecreate e2e tests

Signed-off-by: Max Cao <[email protected]>

Author: maxcao13

vertical-pod-autoscaler/e2e/go.mod

@@ -14,7 +14,7 @@ require (
 	k8s.io/apimachinery v0.32.0
 	k8s.io/autoscaler/vertical-pod-autoscaler v1.2.1
 	k8s.io/client-go v0.32.0
-	k8s.io/component-base v0.32.0
+	k8s.io/component-base v0.32.2
 	k8s.io/klog/v2 v2.130.1
 	k8s.io/kubernetes v1.32.0
 	k8s.io/pod-security-admission v0.32.0

vertical-pod-autoscaler/e2e/go.sum

@@ -92,6 +92,8 @@ github.com/gogo/protobuf v1.3.2 h1:Ov1cvc58UF3b5XjBnZv7+opcTcQFZebYjWzi34vdm4Q=
 github.com/gogo/protobuf v1.3.2/go.mod h1:P1XiOD3dCwIKUDQYPy72D8LYyHL2YPYrpS2s69NZV8Q=
 github.com/golang-jwt/jwt/v4 v4.5.0 h1:7cYmW1XlMY7h7ii7UhUyChSgS5wUJEnm9uZVTGqOWzg=
 github.com/golang-jwt/jwt/v4 v4.5.0/go.mod h1:m21LjoU+eqJr34lmDMbreY2eSTRJ1cv77w39/MY0Ch0=
+github.com/golang/mock v1.6.0 h1:ErTB+efbowRARo13NNdxyJji2egdxLGQhRaY+DUumQc=
+github.com/golang/mock v1.6.0/go.mod h1:p6yTPP+5HYm5mzsMV8JkE6ZKdX+/wYM6Hr+LicevLPs=
 github.com/golang/protobuf v1.5.4 h1:i7eJL8qZTpSEXOPTxNKhASYpMn+8e5Q6AdndVa1dWek=
 github.com/golang/protobuf v1.5.4/go.mod h1:lnTiLA8Wa4RWRcIUkrtSVa5nRhsEGBg48fD6rSs7xps=
 github.com/google/btree v1.0.1 h1:gK4Kx5IaGY9CD5sPJ36FHiBJ6ZXl0kilRiiCj+jdYp4=

vertical-pod-autoscaler/e2e/v1/actuation.go

@@ -35,6 +35,7 @@ import (
 	"k8s.io/apimachinery/pkg/util/wait"
 	"k8s.io/autoscaler/vertical-pod-autoscaler/e2e/utils"
 	vpa_types "k8s.io/autoscaler/vertical-pod-autoscaler/pkg/apis/autoscaling.k8s.io/v1"
+	restriction "k8s.io/autoscaler/vertical-pod-autoscaler/pkg/updater/restriction"
 	"k8s.io/autoscaler/vertical-pod-autoscaler/pkg/utils/annotations"
 	clientset "k8s.io/client-go/kubernetes"
 	"k8s.io/kubernetes/test/e2e/framework"
@@ -50,6 +51,237 @@ import (
 	"github.com/onsi/gomega"
 )
 
+var _ = ActuationSuiteE2eDescribe("Actuation [InPlaceOrRecreate]", func() {
+	f := framework.NewDefaultFramework("vertical-pod-autoscaling")
+	f.NamespacePodSecurityEnforceLevel = podsecurity.LevelBaseline
+
+	ginkgo.BeforeEach(func() {
+		checkInPlaceOrRecreateTestsEnabled(f, true, true)
+	})
+
+	ginkgo.It("still applies recommendations on restart when update mode is InPlaceOrRecreate", func() {
+		ginkgo.By("Setting up a hamster deployment")
+		SetupHamsterDeployment(f, "100m", "100Mi", defaultHamsterReplicas)
+		podList, err := GetHamsterPods(f)
+		gomega.Expect(err).NotTo(gomega.HaveOccurred())
+		podSet := MakePodSet(podList)
+
+		ginkgo.By("Setting up a VPA CRD in mode InPlaceOrRecreate")
+		containerName := GetHamsterContainerNameByIndex(0)
+		vpaCRD := test.VerticalPodAutoscaler().
+			WithName("hamster-vpa").
+			WithNamespace(f.Namespace.Name).
+			WithTargetRef(hamsterTargetRef).
+			WithUpdateMode(vpa_types.UpdateModeInPlaceOrRecreate).
+			WithContainer(containerName).
+			AppendRecommendation(
+				test.Recommendation().
+					WithContainer(containerName).
+					WithTarget("200m", "").
+					WithLowerBound("200m", "").
+					WithUpperBound("200m", "").
+					GetContainerResources()).
+			Get()
+
+		InstallVPA(f, vpaCRD)
+		updatedCPURequest := ParseQuantityOrDie("200m")
+
+		ginkgo.By(fmt.Sprintf("Waiting for pods to be evicted, hoping it won't happen, sleep for %s", VpaEvictionTimeout.String()))
+		CheckNoPodsEvicted(f, podSet)
+		ginkgo.By("Forcefully killing one pod")
+		killPod(f, podList)
+
+		ginkgo.By("Checking that request was modified after forceful restart")
+		updatedPodList, _ := GetHamsterPods(f)
+		var foundUpdated int32
+		for _, pod := range updatedPodList.Items {
+			podRequest := getCPURequest(pod.Spec)
+			framework.Logf("podReq: %v", podRequest)
+			if podRequest.Cmp(updatedCPURequest) == 0 {
+				foundUpdated += 1
+			}
+		}
+		gomega.Expect(foundUpdated).To(gomega.Equal(defaultHamsterReplicas))
+	})
+
+	ginkgo.It("applies in-place updates to all containers when update mode is InPlaceOrRecreate", func() {
+		ginkgo.By("Setting up a hamster deployment")
+		d := NewNHamstersDeployment(f, 2 /*number of containers*/)
+		d.Spec.Template.Spec.Containers[0].Resources.Requests = apiv1.ResourceList{
+			apiv1.ResourceCPU:    ParseQuantityOrDie("100m"),
+			apiv1.ResourceMemory: ParseQuantityOrDie("100Mi"),
+		}
+		d.Spec.Template.Spec.Containers[1].Resources.Requests = apiv1.ResourceList{
+			apiv1.ResourceCPU:    ParseQuantityOrDie("100m"),
+			apiv1.ResourceMemory: ParseQuantityOrDie("100Mi"),
+		}
+		targetCPU := "200m"
+		_ = startDeploymentPods(f, d) // 3 replicas
+		container1Name := GetHamsterContainerNameByIndex(0)
+		container2Name := GetHamsterContainerNameByIndex(1)
+		podList, err := GetHamsterPods(f)
+		gomega.Expect(err).NotTo(gomega.HaveOccurred())
+
+		ginkgo.By("Setting up a VPA CRD")
+		vpaCRD := test.VerticalPodAutoscaler().
+			WithName("hamster-vpa").
+			WithNamespace(f.Namespace.Name).
+			WithTargetRef(hamsterTargetRef).
+			WithContainer(container1Name).
+			WithContainer(container2Name).
+			WithUpdateMode(vpa_types.UpdateModeInPlaceOrRecreate).
+			AppendRecommendation(
+				test.Recommendation().
+					WithContainer(container1Name).
+					WithTarget(targetCPU, "").
+					WithLowerBound("200m", "").
+					WithUpperBound("200m", "").
+					GetContainerResources()).
+			AppendRecommendation(
+				test.Recommendation().
+					WithContainer(container2Name).
+					WithTarget(targetCPU, "").
+					WithLowerBound("200m", "").
+					WithUpperBound("200m", "").
+					GetContainerResources()).
+			Get()
+
+		InstallVPA(f, vpaCRD)
+
+		ginkgo.By("Checking that resources were modified due to in-place update, not due to evictions")
+		err = WaitForPodsUpdatedWithoutEviction(f, podList)
+		gomega.Expect(err).NotTo(gomega.HaveOccurred())
+
+		ginkgo.By("Checking that container resources were actually updated")
+		gomega.Eventually(func() error {
+			updatedPodList, err := GetHamsterPods(f)
+			if err != nil {
+				return err
+			}
+			for _, pod := range updatedPodList.Items {
+				for _, container := range pod.Spec.Containers {
+					if container.Name == container1Name {
+						request := container.Resources.Requests[apiv1.ResourceCPU]
+						if request.Cmp(ParseQuantityOrDie(targetCPU)) != 0 {
+							framework.Logf("%v/%v has not been updated to %v yet: currently=%v", pod.Name, container.Name, targetCPU, request.String())
+							return fmt.Errorf("%s CPU request not updated", container1Name)
+						}
+					}
+					if container.Name == container2Name {
+						request := container.Resources.Requests[apiv1.ResourceCPU]
+						if request.Cmp(ParseQuantityOrDie(targetCPU)) != 0 {
+							framework.Logf("%v/%v has not been updated to %v yet: currently=%v", pod.Name, container.Name, targetCPU, request.String())
+							return fmt.Errorf("%s CPU request not updated", container2Name)
+						}
+					}
+				}
+			}
+			return nil
+		}, VpaInPlaceTimeout*3, 15*time.Second).Should(gomega.Succeed())
+	})
+
+	ginkgo.It("falls back to evicting pods when in-place update is Infeasible when update mode is InPlaceOrRecreate", func() {
+		ginkgo.By("Setting up a hamster deployment")
+		replicas := int32(2)
+		SetupHamsterDeployment(f, "100m", "100Mi", replicas)
+		podList, err := GetHamsterPods(f)
+		gomega.Expect(err).NotTo(gomega.HaveOccurred())
+
+		ginkgo.By("Setting up a VPA CRD")
+		containerName := GetHamsterContainerNameByIndex(0)
+		updatedCPU := "999" // infeasible target
+		vpaCRD := test.VerticalPodAutoscaler().
+			WithName("hamster-vpa").
+			WithNamespace(f.Namespace.Name).
+			WithTargetRef(hamsterTargetRef).
+			WithContainer(containerName).
+			WithUpdateMode(vpa_types.UpdateModeInPlaceOrRecreate).
+			AppendRecommendation(
+				test.Recommendation().
+					WithContainer(containerName).
+					WithTarget(updatedCPU, "").
+					WithLowerBound("200m", "").
+					WithUpperBound("200m", "").
+					GetContainerResources()).
+			Get()
+
+		InstallVPA(f, vpaCRD)
+
+		ginkgo.By("Waiting for pods to be evicted")
+		err = WaitForPodsEvicted(f, podList)
+		gomega.Expect(err).NotTo(gomega.HaveOccurred())
+	})
+
+	ginkgo.It("falls back to evicting pods when resize is Deferred and more than 5 minute has elapsed since last in-place update when update mode is InPlaceOrRecreate", func() {
+		ginkgo.By("Setting up a hamster deployment")
+		replicas := int32(2)
+		SetupHamsterDeployment(f, "100m", "100Mi", replicas)
+		podList, err := GetHamsterPods(f)
+		gomega.Expect(err).NotTo(gomega.HaveOccurred())
+
+		ginkgo.By("Setting up a VPA CRD")
+		containerName := GetHamsterContainerNameByIndex(0)
+
+		// get node name
+		nodeName := podList.Items[0].Spec.NodeName
+		node, err := f.ClientSet.CoreV1().Nodes().Get(context.TODO(), nodeName, metav1.GetOptions{})
+		gomega.Expect(err).NotTo(gomega.HaveOccurred())
+		allocatableCPU := node.Status.Allocatable[apiv1.ResourceCPU]
+		updatedCPU := allocatableCPU.String()
+
+		vpaCRD := test.VerticalPodAutoscaler().
+			WithName("hamster-vpa").
+			WithNamespace(f.Namespace.Name).
+			WithTargetRef(hamsterTargetRef).
+			WithContainer(containerName).
+			WithUpdateMode(vpa_types.UpdateModeInPlaceOrRecreate).
+			AppendRecommendation(
+				test.Recommendation().
+					WithContainer(containerName).
+					WithTarget(updatedCPU, "").
+					WithLowerBound("200m", "").
+					WithUpperBound("200m", "").
+					GetContainerResources()).
+			Get()
+
+		InstallVPA(f, vpaCRD)
+
+		ginkgo.By("Waiting for status to be Deferred")
+		gomega.Eventually(func() error {
+			updatedPodList, err := GetHamsterPods(f)
+			if err != nil {
+				return err
+			}
+			for _, pod := range updatedPodList.Items {
+				if pod.Status.Resize == apiv1.PodResizeStatusDeferred {
+					return nil
+				}
+			}
+			return fmt.Errorf("status not deferred")
+		}, VpaInPlaceTimeout, 5*time.Second).Should(gomega.Succeed())
+
+		ginkgo.By("Making sure pods are not evicted yet")
+		gomega.Consistently(func() error {
+			updatedPodList, err := GetHamsterPods(f)
+			if err != nil {
+				return fmt.Errorf("failed to get pods: %v", err)
+			}
+			for _, pod := range updatedPodList.Items {
+				request := getCPURequestFromStatus(pod.Status)
+				if request.Cmp(ParseQuantityOrDie(updatedCPU)) == 0 {
+					framework.Logf("%v/%v updated to %v, that wasn't supposed to happen this early", pod.Name, containerName, updatedCPU)
+					return fmt.Errorf("%s CPU request should not have been updated", containerName)
+				}
+			}
+			return nil
+		}, restriction.DeferredResizeUpdateTimeout, 10*time.Second).Should(gomega.Succeed())
+
+		ginkgo.By("Waiting for pods to be evicted")
+		err = WaitForPodsEvicted(f, podList)
+		gomega.Expect(err).NotTo(gomega.HaveOccurred())
+	})
+})
+
 var _ = ActuationSuiteE2eDescribe("Actuation", func() {
 	f := framework.NewDefaultFramework("vertical-pod-autoscaling")
 	f.NamespacePodSecurityEnforceLevel = podsecurity.LevelBaseline
@@ -519,6 +751,10 @@ func getCPURequest(podSpec apiv1.PodSpec) resource.Quantity {
 	return podSpec.Containers[0].Resources.Requests[apiv1.ResourceCPU]
 }
 
+func getCPURequestFromStatus(podStatus apiv1.PodStatus) resource.Quantity {
+	return podStatus.ContainerStatuses[0].Resources.Requests[apiv1.ResourceCPU]
+}
+
 func killPod(f *framework.Framework, podList *apiv1.PodList) {
 	f.ClientSet.CoreV1().Pods(f.Namespace.Name).Delete(context.TODO(), podList.Items[0].Name, metav1.DeleteOptions{})
 	err := WaitForPodsRestarted(f, podList)

vertical-pod-autoscaler/e2e/v1/admission_controller.go

@@ -37,10 +37,19 @@ import (
 	"github.com/onsi/gomega"
 )
 
+const (
+	webhookConfigName = "vpa-webhook-config"
+	webhookName       = "vpa.k8s.io"
+)
+
 var _ = AdmissionControllerE2eDescribe("Admission-controller", func() {
 	f := framework.NewDefaultFramework("vertical-pod-autoscaling")
 	f.NamespacePodSecurityEnforceLevel = podsecurity.LevelBaseline
 
+	ginkgo.BeforeEach(func() {
+		waitForVpaWebhookRegistration(f)
+	})
+
 	ginkgo.It("starts pods with new recommended request", func() {
 		d := NewHamsterDeploymentWithResources(f, ParseQuantityOrDie("100m") /*cpu*/, ParseQuantityOrDie("100Mi") /*memory*/)
 
@@ -908,6 +917,40 @@ var _ = AdmissionControllerE2eDescribe("Admission-controller", func() {
 		gomega.Expect(err.Error()).To(gomega.MatchRegexp(`.*admission webhook .*vpa.* denied the request: .*`), "Admission controller did not inspect the object")
 	})
 
+	ginkgo.It("starts pods with new recommended request with InPlaceOrRecreate mode", func() {
+		checkInPlaceOrRecreateTestsEnabled(f, true, false)
+
+		d := NewHamsterDeploymentWithResources(f, ParseQuantityOrDie("100m") /*cpu*/, ParseQuantityOrDie("100Mi") /*memory*/)
+
+		ginkgo.By("Setting up a VPA CRD")
+		containerName := GetHamsterContainerNameByIndex(0)
+		vpaCRD := test.VerticalPodAutoscaler().
+			WithName("hamster-vpa").
+			WithNamespace(f.Namespace.Name).
+			WithTargetRef(hamsterTargetRef).
+			WithContainer(containerName).
+			WithUpdateMode(vpa_types.UpdateModeInPlaceOrRecreate).
+			AppendRecommendation(
+				test.Recommendation().
+					WithContainer(containerName).
+					WithTarget("250m", "200Mi").
+					WithLowerBound("250m", "200Mi").
+					WithUpperBound("250m", "200Mi").
+					GetContainerResources()).
+			Get()
+
+		InstallVPA(f, vpaCRD)
+
+		ginkgo.By("Setting up a hamster deployment")
+		podList := startDeploymentPods(f, d)
+
+		// Originally Pods had 100m CPU, 100Mi of memory, but admission controller
+		// should change it to recommended 250m CPU and 200Mi of memory.
+		for _, pod := range podList.Items {
+			gomega.Expect(pod.Spec.Containers[0].Resources.Requests[apiv1.ResourceCPU]).To(gomega.Equal(ParseQuantityOrDie("250m")))
+			gomega.Expect(pod.Spec.Containers[0].Resources.Requests[apiv1.ResourceMemory]).To(gomega.Equal(ParseQuantityOrDie("200Mi")))
+		}
+	})
 })
 
 func startDeploymentPods(f *framework.Framework, deployment *appsv1.Deployment) *apiv1.PodList {
@@ -962,3 +1005,17 @@ func startDeploymentPods(f *framework.Framework, deployment *appsv1.Deployment)
 	gomega.Expect(err).NotTo(gomega.HaveOccurred(), "when listing pods after deployment resize")
 	return podList
 }
+
+func waitForVpaWebhookRegistration(f *framework.Framework) {
+	ginkgo.By("Waiting for VPA webhook registration")
+	gomega.Eventually(func() bool {
+		webhook, err := f.ClientSet.AdmissionregistrationV1().MutatingWebhookConfigurations().Get(context.TODO(), webhookConfigName, metav1.GetOptions{})
+		if err != nil {
+			return false
+		}
+		if webhook != nil && len(webhook.Webhooks) > 0 && webhook.Webhooks[0].Name == webhookName {
+			return true
+		}
+		return false
+	}, 3*time.Minute, 5*time.Second).Should(gomega.BeTrue(), "Webhook was not registered in the cluster")
+}