Add e2e test for NVIDIA MIG support on Ubuntu 24.04

- Add Test_Ubuntu2404_NvidiaDevicePluginRunning_MIG to validate MIG
  functionality
- Configure test with Standard_NC24ads_A100_v4 VM size and MIG2g instance
  profile
- Add ValidateMIGModeEnabled validator to check MIG mode is enabled via
  nvidia-smi
- Add ValidateMIGInstancesCreated validator to verify MIG instances are properly
  created
- Test validates device plugin, DCGM exporter, and GPU resource scheduling with
  MIG

Signed-off-by: Suraj Deshmukh <[email protected]>

Author: surajssd

e2e/scenario_gpu_managed_experience_test.go

@@ -190,3 +190,65 @@ func Test_AzureLinux3_NvidiaDevicePluginRunning(t *testing.T) {
 		},
 	})
 }
+
+func Test_Ubuntu2404_NvidiaDevicePluginRunning_MIG(t *testing.T) {
+	RunScenario(t, &Scenario{
+		Description: "Tests that NVIDIA device plugin and DCGM Exporter work with MIG enabled on Ubuntu 24.04 GPU nodes",
+		Tags: Tags{
+			GPU: true,
+		},
+		Config: Config{
+			Cluster: ClusterKubenet,
+			VHD:     config.VHDUbuntu2404Gen2Containerd,
+			BootstrapConfigMutator: func(nbc *datamodel.NodeBootstrappingConfiguration) {
+				nbc.AgentPoolProfile.VMSize = "Standard_NC24ads_A100_v4"
+				nbc.ConfigGPUDriverIfNeeded = true
+				nbc.EnableGPUDevicePluginIfNeeded = true
+				nbc.EnableNvidia = true
+				nbc.GPUInstanceProfile = "MIG2g"
+			},
+			VMConfigMutator: func(vmss *armcompute.VirtualMachineScaleSet) {
+				vmss.SKU.Name = to.Ptr("Standard_NC24ads_A100_v4")
+				if vmss.Tags == nil {
+					vmss.Tags = map[string]*string{}
+				}
+				vmss.Tags["EnableManagedGPUExperience"] = to.Ptr("true")
+			},
+			Validator: func(ctx context.Context, s *Scenario) {
+				os := "ubuntu"
+				osVersion := "r2404"
+
+				// Validate that the NVIDIA device plugin binary was installed correctly
+				versions := components.GetExpectedPackageVersions("nvidia-device-plugin", os, osVersion)
+				require.Lenf(s.T, versions, 1, "Expected exactly one nvidia-device-plugin version for %s %s but got %d", os, osVersion, len(versions))
+				ValidateInstalledPackageVersion(ctx, s, "nvidia-device-plugin", versions[0])
+
+				// Validate that the NVIDIA device plugin systemd service is running
+				ValidateNvidiaDevicePluginServiceRunning(ctx, s)
+
+				// Validate that MIG mode is enabled via nvidia-smi
+				ValidateMIGModeEnabled(ctx, s)
+
+				// Validate that MIG instances are created
+				ValidateMIGInstancesCreated(ctx, s, "MIG 2g.20gb")
+
+				// Validate that GPU resources are advertised by the device plugin
+				ValidateNodeAdvertisesGPUResources(ctx, s, 3)
+
+				// Validate that MIG workloads can be scheduled
+				ValidateGPUWorkloadSchedulable(ctx, s, 3)
+
+				// Validate that the NVIDIA DCGM packages were installed correctly
+				for _, packageName := range getDCGMPackageNames(os) {
+					versions := components.GetExpectedPackageVersions(packageName, os, osVersion)
+					require.Lenf(s.T, versions, 1, "Expected exactly one %s version for %s %s but got %d", packageName, os, osVersion, len(versions))
+					ValidateInstalledPackageVersion(ctx, s, packageName, versions[0])
+				}
+
+				ValidateNvidiaDCGMExporterSystemDServiceRunning(ctx, s)
+				ValidateNvidiaDCGMExporterIsScrapable(ctx, s)
+				ValidateNvidiaDCGMExporterScrapeCommonMetric(ctx, s)
+			},
+		},
+	})
+}

e2e/validators.go

@@ -1182,3 +1182,39 @@ func ValidateNvidiaDCGMExporterScrapeCommonMetric(ctx context.Context, s *Scenar
 	}
 	execScriptOnVMForScenarioValidateExitCode(ctx, s, strings.Join(command, "\n"), 0, "Nvidia DCGM Exporter is not returning DCGM_FI_DEV_GPU_UTIL")
 }
+
+func ValidateMIGModeEnabled(ctx context.Context, s *Scenario) {
+	s.T.Helper()
+	s.T.Logf("validating that MIG mode is enabled")
+
+	command := []string{
+		"set -ex",
+		// Grep to verify it contains 'Enabled' - this will fail if MIG is disabled
+		"sudo nvidia-smi --query-gpu=mig.mode.current --format=csv,noheader | grep -i 'Enabled'",
+	}
+	execResult := execScriptOnVMForScenarioValidateExitCode(ctx, s, strings.Join(command, "\n"), 0, "MIG mode is not enabled")
+
+	stdout := strings.TrimSpace(execResult.stdout.String())
+	s.T.Logf("MIG mode status: %s", stdout)
+	require.Contains(s.T, stdout, "Enabled", "expected MIG mode to be enabled, but got: %s", stdout)
+	s.T.Logf("MIG mode is enabled")
+}
+
+func ValidateMIGInstancesCreated(ctx context.Context, s *Scenario, migProfile string) {
+	s.T.Helper()
+	s.T.Logf("validating that MIG instances are created with profile %s", migProfile)
+
+	command := []string{
+		"set -ex",
+		// List MIG devices using nvidia-smi
+		"sudo nvidia-smi mig -lgi",
+		// Ensure the output contains the expected MIG profile (will fail if "No MIG-enabled devices found")
+		"sudo nvidia-smi mig -lgi | grep -v 'No MIG-enabled devices found' | grep -q '" + migProfile + "'",
+	}
+	execResult := execScriptOnVMForScenarioValidateExitCode(ctx, s, strings.Join(command, "\n"), 0, "MIG instances with profile "+migProfile+" were not found")
+
+	stdout := execResult.stdout.String()
+	require.Contains(s.T, stdout, migProfile, "expected to find MIG profile %s in output, but did not.\nOutput:\n%s", migProfile, stdout)
+	require.NotContains(s.T, stdout, "No MIG-enabled devices found", "no MIG devices were created.\nOutput:\n%s", stdout)
+	s.T.Logf("MIG instances with profile %s are created", migProfile)
+}