[WIP] DRA e2e test to validate Nvidia gpu

test/extended/dra/common_spec.go

@@ -0,0 +1,117 @@
+package dra
+
+import (
+	"context"
+	"fmt"
+
+	g "github.com/onsi/ginkgo/v2"
+	o "github.com/onsi/gomega"
+
+	drae2eutility "github.com/openshift/origin/test/extended/dra/utility"
+	exutil "github.com/openshift/origin/test/extended/util"
+
+	corev1 "k8s.io/api/core/v1"
+	resourceapi "k8s.io/api/resource/v1beta1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/kubernetes/test/e2e/framework"
+	e2epodutil "k8s.io/kubernetes/test/e2e/framework/pod"
+	"k8s.io/utils/ptr"
+)
+
+type driver interface {
+	DeviceClassName() string
+}
+
+type commonSpec struct {
+	f  *framework.Framework
+	oc *exutil.CLI
+
+	driver       driver
+	newContainer func(name string) corev1.Container
+	// the node onto which the pod is expected to run
+	node *corev1.Node
+}
+
+func (spec commonSpec) Test(t g.GinkgoTInterface) {
+	// create a namespace
+	ns, err := spec.f.CreateNamespace(context.TODO(), "dra", nil)
+	framework.ExpectNoError(err)
+
+	deviceclass := spec.driver.DeviceClassName()
+	helper := drae2eutility.NewHelper(ns.Name, spec.driver.DeviceClassName())
+
+	claim := &resourceapi.ResourceClaimTemplate{
+		ObjectMeta: metav1.ObjectMeta{
+			Namespace: ns.Name,
+			Name:      "single-gpu",
+		},
+		Spec: resourceapi.ResourceClaimTemplateSpec{
+			Spec: resourceapi.ResourceClaimSpec{
+				Devices: resourceapi.DeviceClaim{
+					Requests: []resourceapi.DeviceRequest{
+						{
+							Name:            "gpu",
+							DeviceClassName: deviceclass,
+						},
+					},
+				},
+			},
+		},
+	}
+
+	// one pod, one container
+	pod := helper.NewPod("pod")
+	ctr := spec.newContainer("ctr")
+	ctr.Resources.Claims = []corev1.ResourceClaim{{Name: "gpu"}}
+	pod.Spec.Containers = append(pod.Spec.Containers, ctr)
+	pod.Spec.ResourceClaims = []corev1.PodResourceClaim{
+		{
+			Name:                      "gpu",
+			ResourceClaimTemplateName: ptr.To(claim.Name),
+		},
+	}
+	pod.Spec.Tolerations = []corev1.Toleration{
+		{
+			Key:      "nvidia.com/gpu",
+			Operator: corev1.TolerationOpExists,
+			Effect:   corev1.TaintEffectNoSchedule,
+		},
+	}
+
+	g.By("creating external claim and pod")
+	resource := spec.f.ClientSet.ResourceV1beta1()
+	claim, err = resource.ResourceClaimTemplates(ns.Name).Create(context.Background(), claim, metav1.CreateOptions{})
+	o.Expect(err).To(o.BeNil())
+
+	core := spec.f.ClientSet.CoreV1()
+	pod, err = core.Pods(ns.Name).Create(context.Background(), pod, metav1.CreateOptions{})
+	o.Expect(err).To(o.BeNil())
+
+	g.By(fmt.Sprintf("waiting for pod %s/%s to be running", pod.Namespace, pod.Name))
+	err = e2epodutil.WaitForPodRunningInNamespace(context.Background(), spec.f.ClientSet, pod)
+	o.Expect(err).To(o.BeNil())
+
+	pod, err = core.Pods(ns.Name).Get(context.Background(), pod.Name, metav1.GetOptions{})
+	o.Expect(err).To(o.BeNil())
+	o.Expect(pod.Spec.NodeName).To(o.Equal(spec.node.Name))
+
+	result, err := resource.ResourceClaims(ns.Name).List(context.Background(), metav1.ListOptions{})
+	o.Expect(err).To(o.BeNil())
+	o.Expect(len(result.Items)).To(o.Equal(1))
+	rc := result.Items[0]
+	o.Expect(rc.Status.Allocation).NotTo(o.BeNil())
+	o.Expect(len(rc.Status.Allocation.Devices.Results)).To(o.Equal(1))
+
+	allocation := rc.Status.Allocation.Devices.Results[0]
+	o.Expect(allocation.Request).To(o.Equal("gpu"))
+	o.Expect(allocation.Driver).To(o.Equal(deviceclass))
+	o.Expect(allocation.Device).To(o.Equal("gpu-0"))
+
+	g.By(fmt.Sprintf("retrieving logs for pod %s/%s", pod.Namespace, pod.Name))
+	args := []string{
+		"-n", pod.Namespace, pod.Name, "--all-containers",
+	}
+	logs, err := spec.oc.AsAdmin().Run("logs").Args(args...).Output()
+	o.Expect(err).To(o.BeNil())
+	t.Logf("\n%s\n", logs)
+}