Enable labels for ClusterUUID and CliqueId

Signed-off-by: Carlos Eduardo Arango Gutierrez <[email protected]>

deployments/helm/nvidia-device-plugin/templates/daemonset-gfd.yml

@@ -182,6 +182,10 @@ spec:
             mountPath: "/etc/kubernetes/node-feature-discovery/features.d"
           - name: host-sys
             mountPath: "/sys"
+        {{- if .Values.imexEnabled }}
+          - name: imex-nodes-config
+            mountPath: "/etc/nvidia-imex/nodes_config.cfg"
+        {{- end }}
         {{- if $options.hasConfigMap }}
           - name: available-configs
             mountPath: /available-configs
@@ -199,6 +203,11 @@ spec:
         - name: host-sys
           hostPath:
             path: "/sys"
+      {{- if .Values.imexEnabled }}
+        - name: imex-nodes-config
+          hostPath:
+            path: "/etc/nvidia-imex/nodes_config.cfg"
+      {{- end }}
       {{- if $options.hasConfigMap }}
         - name: available-configs
           configMap:

deployments/helm/nvidia-device-plugin/values.yaml

@@ -35,6 +35,7 @@ deviceIDStrategy: null
 nvidiaDriverRoot: null
 gdsEnabled: null
 mofedEnabled: null
+imexEnabled: false
 deviceDiscoveryStrategy: null
 
 nameOverride: ""

internal/lm/nvml.go

@@ -17,8 +17,12 @@
 package lm
 
 import (
+	"bufio"
 	"errors"
 	"fmt"
+	"math/rand"
+	"os"
+	"sort"
 	"strconv"
 	"strings"
 
@@ -28,6 +32,7 @@ import (
 
 	spec "github.com/NVIDIA/k8s-device-plugin/api/config/v1"
 	"github.com/NVIDIA/k8s-device-plugin/internal/resource"
+	"github.com/google/uuid"
 )
 
 var errMPSSharingNotSupported = errors.New("MPS sharing is not supported")
@@ -80,13 +85,22 @@ func NewDeviceLabeler(manager resource.Manager, config *spec.Config) (Labeler, e
 		return nil, fmt.Errorf("error creating resource labeler: %v", err)
 	}
 
+	var imexLabeler Labeler
+	if *config.Flags.GFD.ImexNodesConfig != "" {
+		imexLabeler, err = newImexDomainLabeler(*config.Flags.GFD.ImexNodesConfig)
+		if err != nil {
+			return nil, fmt.Errorf("error creating imex domain labeler: %v", err)
+		}
+	}
+
 	l := Merge(
 		machineTypeLabeler,
 		versionLabeler,
 		migCapabilityLabeler,
 		sharingLabeler,
 		resourceLabeler,
 		gpuModeLabeler,
+		imexLabeler,
 	)
 
 	return l, nil
@@ -218,6 +232,41 @@ func newGPUModeLabeler(devices []resource.Device) (Labeler, error) {
 	return labels, nil
 }
 
+func newImexDomainLabeler(configFile string) (Labeler, error) {
+	// read file and parse it
+	imexConfig, err := os.Open(configFile)
+	if err != nil {
+		return nil, fmt.Errorf("failed to read imex config file: %v", err)
+	}
+	defer imexConfig.Close()
+
+	// Read the file line by line
+	var ips []string
+	scanner := bufio.NewScanner(imexConfig)
+	for scanner.Scan() {
+		line := scanner.Text()
+		ips = append(ips, line)
+	}
+
+	if err := scanner.Err(); err != nil {
+		return nil, fmt.Errorf("failed to read imex config file: %v", err)
+	}
+
+	// Sort the IP addresses
+	sort.Strings(ips)
+
+	// Join the sorted IPs into a single string
+	sortedIPs := strings.Join(ips, "\n")
+
+	hashedconfig := generateUUIDs(sortedIPs, 1)[0]
+
+	labels := Labels{
+		"nvidia.com/gpu.imex-domain": hashedconfig,
+	}
+
+	return labels, nil
+}
+
 func getModeForClasses(classes []uint32) string {
 	if len(classes) == 0 {
 		return "unknown"
@@ -254,3 +303,25 @@ func getDeviceClasses(devices []resource.Device) ([]uint32, error) {
 	}
 	return classes, nil
 }
+
+func generateUUIDs(seed string, count int) []string {
+	rand := rand.New(rand.NewSource(hash(seed)))
+
+	uuids := make([]string, count)
+	for i := 0; i < count; i++ {
+		charset := make([]byte, 16)
+		rand.Read(charset)
+		uuid, _ := uuid.FromBytes(charset)
+		uuids[i] = uuid.String()
+	}
+
+	return uuids
+}
+
+func hash(s string) int64 {
+	h := int64(0)
+	for _, c := range s {
+		h = 31*h + int64(c)
+	}
+	return h
+}