[CONTP-499] Parsing GPU tags on kubeapiserver collector

comp/core/workloadmeta/collectors/internal/kubelet/kubelet.go

@@ -22,6 +22,7 @@ import (
 	"github.com/DataDog/datadog-agent/pkg/errors"
 	"github.com/DataDog/datadog-agent/pkg/util/containers"
 	pkgcontainersimage "github.com/DataDog/datadog-agent/pkg/util/containers/image"
+	"github.com/DataDog/datadog-agent/pkg/util/gpu"
 	"github.com/DataDog/datadog-agent/pkg/util/kubernetes"
 	"github.com/DataDog/datadog-agent/pkg/util/kubernetes/kubelet"
 	"github.com/DataDog/datadog-agent/pkg/util/log"
@@ -418,21 +419,6 @@ func extractEnvFromSpec(envSpec []kubelet.EnvVar) map[string]string {
 	return env
 }
 
-func extractGPUVendor(gpuNamePrefix kubelet.ResourceName) string {
-	gpuVendor := ""
-	switch gpuNamePrefix {
-	case kubelet.ResourcePrefixNvidiaMIG, kubelet.ResourceGenericNvidiaGPU:
-		gpuVendor = "nvidia"
-	case kubelet.ResourcePrefixAMDGPU:
-		gpuVendor = "amd"
-	case kubelet.ResourcePrefixIntelGPU:
-		gpuVendor = "intel"
-	default:
-		gpuVendor = string(gpuNamePrefix)
-	}
-	return gpuVendor
-}
-
 func extractResources(spec *kubelet.ContainerSpec) workloadmeta.ContainerResources {
 	resources := workloadmeta.ContainerResources{}
 	if cpuReq, found := spec.Resources.Requests[kubelet.ResourceCPU]; found {
@@ -444,24 +430,14 @@ func extractResources(spec *kubelet.ContainerSpec) workloadmeta.ContainerResourc
 	}
 
 	// extract GPU resource info from the possible GPU sources
-	uniqueGPUVendor := make(map[string]bool)
-
-	resourceKeys := make([]kubelet.ResourceName, 0, len(spec.Resources.Requests))
+	uniqueGPUVendor := make(map[string]struct{})
 	for resourceName := range spec.Resources.Requests {
-		resourceKeys = append(resourceKeys, resourceName)
-	}
-
-	for _, gpuResourceName := range kubelet.GetGPUResourceNames() {
-		for _, resourceKey := range resourceKeys {
-			if strings.HasPrefix(string(resourceKey), string(gpuResourceName)) {
-				if gpuReq, found := spec.Resources.Requests[resourceKey]; found {
-					resources.GPURequest = pointer.Ptr(uint64(gpuReq.Value()))
-					uniqueGPUVendor[extractGPUVendor(gpuResourceName)] = true
-					break
-				}
-			}
+		gpuName, found := gpu.ExtractSimpleGPUName(gpu.ResourceGPU(resourceName))
+		if found {
+			uniqueGPUVendor[gpuName] = struct{}{}
 		}
 	}
+
 	gpuVendorList := make([]string, 0, len(uniqueGPUVendor))
 	for GPUVendor := range uniqueGPUVendor {
 		gpuVendorList = append(gpuVendorList, GPUVendor)

comp/core/workloadmeta/collectors/util/kubernetes_resource_parsers/pod.go

@@ -13,6 +13,7 @@ import (
 	corev1 "k8s.io/api/core/v1"
 
 	workloadmeta "github.com/DataDog/datadog-agent/comp/core/workloadmeta/def"
+	"github.com/DataDog/datadog-agent/pkg/util/gpu"
 )
 
 type podParser struct {
@@ -62,6 +63,20 @@ func (p podParser) Parse(obj interface{}) workloadmeta.Entity {
 		rtcName = *pod.Spec.RuntimeClassName
 	}
 
+	var gpuVendorList []string
+	uniqueGPUVendor := make(map[string]struct{})
+	for _, container := range pod.Spec.Containers {
+		for resourceName := range container.Resources.Limits {
+			gpuName, found := gpu.ExtractSimpleGPUName(gpu.ResourceGPU(resourceName))
+			if found {
+				uniqueGPUVendor[gpuName] = struct{}{}
+			}
+		}
+	}
+	for gpuVendor := range uniqueGPUVendor {
+		gpuVendorList = append(gpuVendorList, gpuVendor)
+	}
+
 	return &workloadmeta.KubernetesPod{
 		EntityID: workloadmeta.EntityID{
 			Kind: workloadmeta.KindKubernetesPod,
@@ -81,6 +96,7 @@ func (p podParser) Parse(obj interface{}) workloadmeta.Entity {
 		PriorityClass:              pod.Spec.PriorityClassName,
 		QOSClass:                   string(pod.Status.QOSClass),
 		RuntimeClass:               rtcName,
+		GPUVendorList:              gpuVendorList,
 
 		// Containers could be generated by this collector, but
 		// currently it's not to save on memory, since this is supposed

comp/core/workloadmeta/collectors/util/kubernetes_resource_parsers/pod_test.go

@@ -12,6 +12,7 @@ import (
 
 	"github.com/stretchr/testify/assert"
 	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 
 	workloadmeta "github.com/DataDog/datadog-agent/comp/core/workloadmeta/def"
@@ -54,6 +55,28 @@ func TestPodParser_Parse(t *testing.T) {
 					},
 				},
 			},
+			Containers: []corev1.Container{
+				{
+					Name: "gpuContainer1",
+					Resources: corev1.ResourceRequirements{
+						Limits: corev1.ResourceList{
+							"nvidia.com/gpu": resource.Quantity{
+								Format: "1",
+							},
+						},
+					},
+				},
+				{
+					Name: "gpuContainer2",
+					Resources: corev1.ResourceRequirements{
+						Limits: corev1.ResourceList{
+							"gpu.intel.com/xe": resource.Quantity{
+								Format: "2",
+							},
+						},
+					},
+				},
+			},
 		},
 		Status: corev1.PodStatus{
 			Phase: corev1.PodRunning,
@@ -97,6 +120,7 @@ func TestPodParser_Parse(t *testing.T) {
 		Ready:                      true,
 		IP:                         "127.0.0.1",
 		PriorityClass:              "priorityClass",
+		GPUVendorList:              []string{"nvidia", "intel"},
 		QOSClass:                   "Guaranteed",
 	}
 

go.mod

@@ -123,6 +123,7 @@ replace (
 	github.com/DataDog/datadog-agent/pkg/util/filesystem => ./pkg/util/filesystem
 	github.com/DataDog/datadog-agent/pkg/util/flavor => ./pkg/util/flavor
 	github.com/DataDog/datadog-agent/pkg/util/fxutil => ./pkg/util/fxutil/
+	github.com/DataDog/datadog-agent/pkg/util/gpu => ./pkg/util/gpu
 	github.com/DataDog/datadog-agent/pkg/util/grpc => ./pkg/util/grpc/
 	github.com/DataDog/datadog-agent/pkg/util/hostname/validate => ./pkg/util/hostname/validate/
 	github.com/DataDog/datadog-agent/pkg/util/http => ./pkg/util/http/

pkg/util/gpu/common.go

@@ -0,0 +1,48 @@
+// Unless explicitly stated otherwise all files in this repository are licensed
+// under the Apache License Version 2.0.
+// This product includes software developed at Datadog (https://www.datadoghq.com/).
+// Copyright 2016-present Datadog, Inc.
+
+// Package gpu provides utilities for interacting with GPU resources.
+package gpu
+
+import "strings"
+
+// ResourceGPU represents a GPU resource
+type ResourceGPU string
+
+// Resource name prefixes
+const (
+	gpuNvidiaGeneric ResourceGPU = "nvidia.com/gpu"
+	gpuAMD           ResourceGPU = "amd.com/gpu"
+	gpuIntelXe       ResourceGPU = "gpu.intel.com/xe"
+	gpuInteli915     ResourceGPU = "gpu.intel.com/i915"
+
+	gpuNvidiaMigPrefix ResourceGPU = "nvidia.com/mig"
+)
+
+// longToShortGPUName maps a GPU resource to a simplified name
+var longToShortGPUName = map[ResourceGPU]string{
+	gpuNvidiaGeneric: "nvidia",
+	gpuAMD:           "amd",
+	gpuIntelXe:       "intel",
+	gpuInteli915:     "intel",
+}
+
+// ExtractSimpleGPUName returns a simplified GPU name.
+// If the resource is not recognized, the second return value is false.
+func ExtractSimpleGPUName(gpuName ResourceGPU) (string, bool) {
+	val, ok := longToShortGPUName[gpuName]
+	if ok {
+		return val, true
+	}
+
+	// More complex cases (eg. nvidia.com/mig-3g.20gb => nvidia)
+	switch {
+	case strings.HasPrefix(string(gpuName), string(gpuNvidiaMigPrefix)):
+		return "nvidia", true
+	}
+
+	// Not a GPU resource (or not recognized)
+	return "", false
+}

pkg/util/gpu/common_test.go

@@ -0,0 +1,41 @@
+// Unless explicitly stated otherwise all files in this repository are licensed
+// under the Apache License Version 2.0.
+// This product includes software developed at Datadog (https://www.datadoghq.com/).
+// Copyright 2016-present Datadog, Inc.
+
+package gpu
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestExtractSimpleGPUName(t *testing.T) {
+	tests := []struct {
+		name     string
+		gpuName  ResourceGPU
+		found    bool
+		expected string
+	}{
+		{
+			name:     "known gpu resource",
+			gpuName:  gpuNvidiaGeneric,
+			found:    true,
+			expected: "nvidia",
+		},
+		{
+			name:     "unknown gpu resource",
+			gpuName:  ResourceGPU("cpu"),
+			found:    false,
+			expected: "",
+		},
+	}
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			actual, found := ExtractSimpleGPUName(test.gpuName)
+			assert.Equal(t, test.found, found)
+			assert.Equal(t, test.expected, actual)
+		})
+	}
+}

releasenotes/notes/kubeapiserver-gpu-tagging-e6202bc782982e5d.yaml

@@ -0,0 +1,12 @@
+# Each section from every release note are combined when the
+# CHANGELOG.rst is rendered. So the text needs to be worded so that
+# it does not depend on any information only available in another
+# section. This may mean repeating some details, but each section
+# must be readable independently of the other.
+#
+# Each section note must be formatted as reStructuredText.
+---
+fixes:
+  - |
+    Include `gpu_vendor` pod tags on the Datadog Cluster Agent when
+    enabling datadog.clusterTagger.collectKubernetesTags.