Make ComptueDomains with DNS names the default

Signed-off-by: Kevin Klues <[email protected]>

Author: klueska

cmd/compute-domain-kubelet-plugin/device_state.go

@@ -25,12 +25,14 @@ import (
 	resourceapi "k8s.io/api/resource/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/util/version"
 	"k8s.io/dynamic-resource-allocation/kubeletplugin"
 	"k8s.io/klog/v2"
 	"k8s.io/kubernetes/pkg/kubelet/checkpointmanager"
 	cdiapi "tags.cncf.io/container-device-interface/pkg/cdi"
 
 	configapi "github.com/NVIDIA/k8s-dra-driver-gpu/api/nvidia.com/resource/v1beta1"
+	"github.com/NVIDIA/k8s-dra-driver-gpu/pkg/featuregates"
 )
 
 type OpaqueDeviceConfig struct {
@@ -62,6 +64,13 @@ func NewDeviceState(ctx context.Context, config *Config) (*DeviceState, error) {
 		return nil, fmt.Errorf("failed to create device library: %w", err)
 	}
 
+	// Check driver version if IMEXDaemonsWithDNSNames feature gate is enabled
+	if featuregates.Enabled(featuregates.IMEXDaemonsWithDNSNames) {
+		if err := validateDriverVersionForIMEXDaemonsWithDNSNames(config.flags, nvdevlib); err != nil {
+			return nil, fmt.Errorf("driver version validation failed: %w", err)
+		}
+	}
+
 	allocatable, err := nvdevlib.enumerateAllPossibleDevices(config)
 	if err != nil {
 		return nil, fmt.Errorf("error enumerating all possible devices: %w", err)
@@ -561,6 +570,32 @@ func (s *DeviceState) getConfigResultsMap(rcs *resourceapi.ResourceClaimStatus,
 	return configResultsMap, nil
 }
 
+// validateDriverVersionForIMEXDaemonsWithDNSNames validates that the driver version
+// meets the minimum requirement for the IMEXDaemonsWithDNSNames feature gate.
+func validateDriverVersionForIMEXDaemonsWithDNSNames(flags *Flags, nvdevlib *deviceLib) error {
+	klog.Infof("Starting driver version validation for IMEXDaemonsWithDNSNames feature...")
+	klog.Infof("Minimum required version: %s", IMEXDaemonsWithDNSNamesMinDriverVersion)
+
+	driverVer, err := nvdevlib.getDriverVersion()
+	if err != nil {
+		return fmt.Errorf("error getting driver version: %w", err)
+	}
+
+	minVersion, err := version.ParseGeneric(IMEXDaemonsWithDNSNamesMinDriverVersion)
+	if err != nil {
+		return fmt.Errorf("error parsing minimum version: %w", err)
+	}
+
+	if driverVer.LessThan(minVersion) {
+		klog.Errorf("IMEXDaemonsWithDNSNames feature requires driver version >= %s, but found %s", minVersion.String(), driverVer.String())
+		klog.Errorf("If installed via helm, set featureGates.IMEXDaemonsWithDNSNames=false to disable feature")
+		return fmt.Errorf("minimum version not satisfied for IMEXDaemonsWithDNSNames feature")
+	}
+
+	klog.Infof("Driver version validation passed: %s >= %s", driverVer.String(), minVersion.String())
+	return nil
+}
+
 // GetOpaqueDeviceConfigs returns an ordered list of the configs contained in possibleConfigs for this driver.
 //
 // Configs can either come from the resource claim itself or from the device

cmd/compute-domain-kubelet-plugin/main.go

@@ -35,8 +35,9 @@ import (
 )
 
 const (
-	DriverName                         = "compute-domain.nvidia.com"
-	DriverPluginCheckpointFileBasename = "checkpoint.json"
+	DriverName                              = "compute-domain.nvidia.com"
+	DriverPluginCheckpointFileBasename      = "checkpoint.json"
+	IMEXDaemonsWithDNSNamesMinDriverVersion = "570.158.01"
 )
 
 type Flags struct {

cmd/compute-domain-kubelet-plugin/nvlib.go

@@ -26,6 +26,7 @@ import (
 	"strconv"
 	"strings"
 
+	"k8s.io/apimachinery/pkg/util/version"
 	"k8s.io/klog/v2"
 	"k8s.io/mount-utils"
 
@@ -125,6 +126,26 @@ func (l deviceLib) alwaysShutdown() {
 	}
 }
 
+func (l deviceLib) getDriverVersion() (*version.Version, error) {
+	if err := l.init(); err != nil {
+		return nil, fmt.Errorf("error initializing NVML: %w", err)
+	}
+	defer l.alwaysShutdown()
+
+	driverVersion, ret := l.nvmllib.SystemGetDriverVersion()
+	if ret != nvml.SUCCESS {
+		return nil, fmt.Errorf("error getting driver version: %v", ret)
+	}
+
+	// Parse the version using Kubernetes version utilities
+	v, err := version.ParseGeneric(driverVersion)
+	if err != nil {
+		return nil, fmt.Errorf("invalid driver version format '%s': %w", driverVersion, err)
+	}
+
+	return v, nil
+}
+
 func (l deviceLib) enumerateAllPossibleDevices(config *Config) (AllocatableDevices, error) {
 	alldevices := make(AllocatableDevices)
 

pkg/featuregates/featuregates.go

@@ -61,8 +61,8 @@ var defaultFeatureGates = map[featuregate.Feature]featuregate.VersionedSpecs{
 	},
 	IMEXDaemonsWithDNSNames: {
 		{
-			Default:    false,
-			PreRelease: featuregate.Alpha,
+			Default:    true,
+			PreRelease: featuregate.Beta,
 			Version:    version.MajorMinor(25, 8),
 		},
 	},