feat: add workload test case for external tests

Signed-off-by: Patrick J.P. Culp <jpculp@amazon.com>

Author: jpculp

test/cases/workload/main_test.go

@@ -0,0 +1,52 @@
+//go:build e2e
+
+package workload
+
+import (
+	"context"
+	"flag"
+	"fmt"
+	"log"
+	"os"
+	"os/signal"
+	"testing"
+
+	"sigs.k8s.io/e2e-framework/pkg/env"
+	"sigs.k8s.io/e2e-framework/pkg/envconf"
+)
+
+const (
+	defaultWorkloadTestTimeoutMinutes = 10
+)
+
+var (
+	testenv               env.Environment
+	workloadTestCommand   *string
+	workloadTestImage     *string
+	workloadTestName      *string
+	workloadTestResources *string
+	workloadTestTimeout   *int
+)
+
+func TestMain(m *testing.M) {
+	workloadTestCommand = flag.String("workloadTestCommand", "", "command for workload test")
+	workloadTestImage = flag.String("workloadTestImage", "", "image for workload test")
+	workloadTestName = flag.String("workloadTestName", "workload-test", "name for workload test")
+	workloadTestResources = flag.String("workloadTestResources", "", "JSON map of resources for workload test (e.g., '{\"nvidia.com/gpu\": \"1\"}')")
+	workloadTestTimeout = flag.Int("workloadTestTimeout", defaultWorkloadTestTimeoutMinutes, fmt.Sprintf("timeout in minutes for workload test (default: %d)", defaultWorkloadTestTimeoutMinutes))
+	cfg, err := envconf.NewFromFlags()
+	if err != nil {
+		log.Fatalf("failed to initialize test environment: %v", err)
+	}
+	testenv = env.NewWithConfig(cfg)
+	ctx, cancel := signal.NotifyContext(context.Background(), os.Interrupt)
+	defer cancel()
+	testenv = testenv.WithContext(ctx)
+
+	testenv.Setup(func(ctx context.Context, config *envconf.Config) (context.Context, error) {
+		log.Println("Starting workload test suite...")
+		return ctx, nil
+	})
+
+	os.Exit(testenv.Run(m))
+}

test/cases/workload/workload_test.go

@@ -0,0 +1,152 @@
+//go:build e2e
+
+package workload
+
+import (
+	"context"
+	"encoding/json"
+	"strings"
+	"testing"
+	"time"
+
+	fwext "github.com/aws/aws-k8s-tester/internal/e2e"
+	"github.com/aws/smithy-go/ptr"
+	batchv1 "k8s.io/api/batch/v1"
+	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"sigs.k8s.io/e2e-framework/klient/wait"
+	"sigs.k8s.io/e2e-framework/pkg/envconf"
+	"sigs.k8s.io/e2e-framework/pkg/features"
+)
+
+func createWorkloadJob(name, image, command string, resources map[string]string, timeoutSeconds int64) *batchv1.Job {
+	job := &batchv1.Job{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      name,
+			Namespace: corev1.NamespaceDefault,
+			Labels:    map[string]string{"app": name},
+		},
+		Spec: batchv1.JobSpec{
+			BackoffLimit: ptr.Int32(4),
+			Template: corev1.PodTemplateSpec{
+				ObjectMeta: metav1.ObjectMeta{
+					Labels: map[string]string{"app": name},
+				},
+				Spec: corev1.PodSpec{
+					RestartPolicy:         corev1.RestartPolicyNever,
+					ActiveDeadlineSeconds: &timeoutSeconds,
+					Containers: []corev1.Container{
+						{
+							Name:            name,
+							Image:           image,
+							Args:            strings.Fields(command),
+							ImagePullPolicy: corev1.PullAlways,
+							Resources:       buildResourceRequirements(resources),
+						},
+					},
+				},
+			},
+		},
+	}
+
+	return job
+}
+
+func buildResourceRequirements(resources map[string]string) corev1.ResourceRequirements {
+	if len(resources) == 0 {
+		return corev1.ResourceRequirements{}
+	}
+	rl := make(corev1.ResourceList)
+	for name, qty := range resources {
+		rl[corev1.ResourceName(name)] = resource.MustParse(qty)
+	}
+	return corev1.ResourceRequirements{Limits: rl, Requests: rl}
+}
+
+func parseResources(resourcesJSON string) (map[string]string, error) {
+	if resourcesJSON == "" {
+		return nil, nil
+	}
+	var resources map[string]string
+	err := json.Unmarshal([]byte(resourcesJSON), &resources)
+	return resources, err
+}
+
+func TestWorkload(t *testing.T) {
+	name := ptr.ToString(workloadTestName)
+	image := ptr.ToString(workloadTestImage)
+	command := ptr.ToString(workloadTestCommand)
+	timeoutMinutes := ptr.ToInt(workloadTestTimeout)
+
+	if name == "" {
+		t.Fatal("workloadTestName must be set to run the test")
+	}
+	if image == "" {
+		t.Fatal("workloadTestImage must be set to run the test")
+	}
+
+	resources, err := parseResources(ptr.ToString(workloadTestResources))
+	if err != nil {
+		t.Fatalf("Failed to parse workloadTestResources: %v", err)
+	}
+
+	feature := features.New(name).WithLabel("suite", "workload")
+	if _, ok := resources["aws.amazon.com/neuron"]; ok {
+		feature = feature.WithLabel("hardware", "neuron")
+	}
+	if _, ok := resources["nvidia.com/gpu"]; ok {
+		feature = feature.WithLabel("hardware", "gpu")
+	}
+
+	workload := feature.Setup(func(ctx context.Context, t *testing.T, cfg *envconf.Config) context.Context {
+		job := createWorkloadJob(name, image, command, resources, int64(timeoutMinutes*60))
+		if len(resources) > 0 {
+			t.Logf("Creating %s job with resources: %v", name, resources)
+		} else {
+			t.Logf("Creating %s job", name)
+		}
+		if err := cfg.Client().Resources().Create(ctx, job); err != nil {
+			t.Fatal(err)
+		}
+		t.Logf("%s job created successfully", name)
+		return ctx
+	}).
+		Assess("Job succeeds", func(ctx context.Context, t *testing.T, cfg *envconf.Config) context.Context {
+			job := &batchv1.Job{
+				ObjectMeta: metav1.ObjectMeta{Name: name, Namespace: corev1.NamespaceDefault},
+			}
+			t.Logf("Waiting for %s job to complete", name)
+			err := wait.For(fwext.NewConditionExtension(cfg.Client().Resources()).JobSucceeded(job),
+				wait.WithContext(ctx),
+				wait.WithTimeout(time.Duration(timeoutMinutes)*time.Minute),
+			)
+			if err != nil {
+				t.Fatal(err)
+			}
+			return ctx
+		}).
+		Teardown(func(ctx context.Context, t *testing.T, cfg *envconf.Config) context.Context {
+			log, err := fwext.GetJobLogs(cfg.Client().RESTConfig(), &batchv1.Job{
+				ObjectMeta: metav1.ObjectMeta{Name: name, Namespace: corev1.NamespaceDefault},
+			})
+			if err != nil {
+				t.Error(err)
+			}
+			t.Logf("Test log for %s:", name)
+			t.Log(log)
+			job := &batchv1.Job{
+				ObjectMeta: metav1.ObjectMeta{Name: name, Namespace: corev1.NamespaceDefault},
+			}
+			if err := cfg.Client().Resources().Delete(ctx, job, func(do *metav1.DeleteOptions) {
+				policy := metav1.DeletePropagationBackground
+				do.PropagationPolicy = &policy
+			}); err != nil {
+				t.Error(err)
+			}
+			return ctx
+		}).
+		Feature()
+
+	testenv.Test(t, workload)
+}