[no-relnote] Move runner declaration to BeforeSuite

Signed-off-by: Carlos Eduardo Arango Gutierrez <eduardoa@nvidia.com>

Author: ArangoGutierrez

tests/e2e/e2e_test.go

@@ -30,6 +30,8 @@ import (
 
 // Test context
 var (
+	runner Runner
+
 	ctx context.Context
 
 	installCTK bool
@@ -49,13 +51,35 @@ func TestMain(t *testing.T) {
 	RegisterFailHandler(Fail)
 
 	ctx = context.Background()
-	getTestEnv()
 
 	RunSpecs(t,
 		suiteName,
 	)
 }
 
+var _ = BeforeSuite(func() {
+	getTestEnv()
+
+	runner = NewRunner(
+		WithHost(sshHost),
+		WithPort(sshPort),
+		WithSshKey(sshKey),
+		WithSshUser(sshUser),
+	)
+
+	if installCTK {
+		installer, err := NewToolkitInstaller(
+			WithRunner(runner),
+			WithImage(imageName+":"+imageTag),
+			WithTemplate(dockerInstallTemplate),
+		)
+		Expect(err).ToNot(HaveOccurred())
+
+		err = installer.Install()
+		Expect(err).ToNot(HaveOccurred())
+	}
+})
+
 // getTestEnv gets the test environment variables
 func getTestEnv() {
 	defer GinkgoRecover()
@@ -73,7 +97,6 @@ func getTestEnv() {
 		sshUser = getRequiredEnvvar[string]("E2E_SSH_USER")
 		sshPort = getEnvVarOrDefault("E2E_SSH_PORT", "22")
 	}
-
 }
 
 // getRequiredEnvvar returns the specified envvar if set or raises an error.

tests/e2e/nvidia-container-cli_test.go

@@ -71,20 +71,12 @@ IN_NS
 
 var _ = Describe("nvidia-container-cli", Ordered, ContinueOnFailure, Label("libnvidia-container"), func() {
 	var (
-		runner                Runner
 		nestedContainerRunner Runner
 		containerName         = "node-container-e2e"
 		hostOutput            string
 	)
 
 	BeforeAll(func(ctx context.Context) {
-		runner = NewRunner(
-			WithHost(sshHost),
-			WithPort(sshPort),
-			WithSshKey(sshKey),
-			WithSshUser(sshUser),
-		)
-
 		var err error
 		nestedContainerRunner, err = NewNestedContainerRunner(runner, installCTK, imageName+":"+imageTag, containerName)
 		Expect(err).ToNot(HaveOccurred())

tests/e2e/nvidia-container-toolkit_test.go

@@ -28,31 +28,11 @@ import (
 
 // Integration tests for Docker runtime
 var _ = Describe("docker", Ordered, ContinueOnFailure, func() {
-	var runner Runner
 	var hostDriverVersion string
 	var hostDriverMajor string
 
 	// Install the NVIDIA Container Toolkit
 	BeforeAll(func(ctx context.Context) {
-		runner = NewRunner(
-			WithHost(sshHost),
-			WithPort(sshPort),
-			WithSshKey(sshKey),
-			WithSshUser(sshUser),
-		)
-
-		if installCTK {
-			installer, err := NewToolkitInstaller(
-				WithRunner(runner),
-				WithImage(imageName+":"+imageTag),
-				WithTemplate(dockerInstallTemplate),
-			)
-			Expect(err).ToNot(HaveOccurred())
-
-			err = installer.Install()
-			Expect(err).ToNot(HaveOccurred())
-		}
-
 		driverOutput, _, err := runner.Run("nvidia-smi -q | grep \"Driver Version\"")
 		Expect(err).ToNot(HaveOccurred())
 		parts := strings.SplitN(driverOutput, ":", 2)

tests/e2e/runner.go

@@ -29,7 +29,7 @@ import (
 )
 
 const (
-	startTestContainerTemplate = `docker run -d --name {{.ContainerName}} --privileged --runtime=nvidia \
+	outerContainerTemplate = `docker run -d --name {{.ContainerName}} --privileged --runtime=nvidia \
     -e NVIDIA_VISIBLE_DEVICES=runtime.nvidia.com/gpu=all \
     -e NVIDIA_DRIVER_CAPABILITIES=all \
 	{{ range $i, $a := .AdditionalArguments -}}
@@ -156,6 +156,8 @@ func NewNestedContainerRunner(runner Runner, installCTK bool, image string, cont
 		return nil, fmt.Errorf("failed to remove container: %w", err)
 	}
 
+	// If installCTK is true, install the toolkit on the host. before creating
+	// the nested container.
 	if installCTK {
 		installer, err := NewToolkitInstaller(
 			WithRunner(runner),
@@ -172,7 +174,6 @@ func NewNestedContainerRunner(runner Runner, installCTK bool, image string, cont
 		}
 	} else {
 		// If installCTK is false, we use the preinstalled toolkit.
-		// TODO: This should be updated for other distributions and other components of the toolkit.
 		output, _, err := runner.Run("ls /lib/**/libnvidia-container*.so.*.*")
 		if err != nil {
 			return nil, fmt.Errorf("failed to list toolkit libraries: %w", err)
@@ -186,16 +187,48 @@ func NewNestedContainerRunner(runner Runner, installCTK bool, image string, cont
 		for _, lib := range strings.Split(output, "\n") {
 			additionalContainerArguments = append(additionalContainerArguments, "-v "+lib+":"+lib)
 		}
-		additionalContainerArguments = append(additionalContainerArguments, "-v /usr/bin/nvidia-container-cli:/usr/bin/nvidia-container-cli")
+
+		// Look for NVIDIA binaries in standard locations and mount them as volumes
+		nvidiaBinaries := []string{
+			"nvidia-container-cli",
+			"nvidia-container-runtime",
+			"nvidia-container-runtime-hook",
+			"nvidia-ctk",
+			"nvidia-cdi-hook",
+			"nvidia-container-runtime.cdi",
+			"nvidia-container-runtime.legacy",
+		}
+
+		searchPaths := []string{
+			"/usr/bin",
+			"/usr/sbin",
+			"/usr/local/bin",
+			"/usr/local/sbin",
+		}
+
+		for _, binary := range nvidiaBinaries {
+			for _, searchPath := range searchPaths {
+				binaryPath := searchPath + "/" + binary
+				// Check if the binary exists at this path
+				checkCmd := fmt.Sprintf("test -f %s && echo 'exists'", binaryPath)
+				output, _, err := runner.Run(checkCmd)
+				if err == nil && strings.TrimSpace(output) == "exists" {
+					// Binary found, add it as a volume mount
+					additionalContainerArguments = append(additionalContainerArguments,
+						fmt.Sprintf("-v %s:%s", binaryPath, binaryPath))
+					break // Move to the next binary once found
+				}
+			}
+		}
 	}
 
 	// Launch the container in detached mode.
-	var startContainerScriptBuilder strings.Builder
-	startContainerTemplate, err := template.New("startContainer").Parse(startTestContainerTemplate)
+	var outerContainerScriptBuilder strings.Builder
+	outerContainerTemplate, err := template.New("outerContainer").Parse(outerContainerTemplate)
 	if err != nil {
 		return nil, fmt.Errorf("failed to parse start container template: %w", err)
 	}
-	err = startContainerTemplate.Execute(&startContainerScriptBuilder, struct {
+	err = outerContainerTemplate.Execute(&outerContainerScriptBuilder, struct {
 		ContainerName       string
 		AdditionalArguments []string
 	}{
@@ -206,8 +239,8 @@ func NewNestedContainerRunner(runner Runner, installCTK bool, image string, cont
 		return nil, fmt.Errorf("failed to execute start container template: %w", err)
 	}
 
-	startContainerScript := startContainerScriptBuilder.String()
-	_, _, err = runner.Run(startContainerScript)
+	outerContainerScript := outerContainerScriptBuilder.String()
+	_, _, err = runner.Run(outerContainerScript)
 	if err != nil {
 		return nil, fmt.Errorf("failed to run start container script: %w", err)
 	}