Add new Tensor Core and GPU capacity metrics

internal/containerinsightscommon/const.go

@@ -83,16 +83,17 @@ const (
 	DiskIOWrite              = "Write"
 	DiskIOTotal              = "Total"
 
-	GpuUtilization    = "gpu_utilization"
-	GpuMemUtilization = "gpu_memory_utilization"
-	GpuMemUsed        = "gpu_memory_used"
-	GpuMemTotal       = "gpu_memory_total"
-	GpuTemperature    = "gpu_temperature"
-	GpuPowerDraw      = "gpu_power_draw"
-	GpuRequest        = "gpu_request"
-	GpuLimit          = "gpu_limit"
-	GpuTotal          = "gpu_total"
-	GpuUniqueId       = "UUID"
+	GpuUtilization           = "gpu_utilization"
+	GpuMemUtilization        = "gpu_memory_utilization"
+	GpuMemUsed               = "gpu_memory_used"
+	GpuMemTotal              = "gpu_memory_total"
+	GpuTemperature           = "gpu_temperature"
+	GpuPowerDraw             = "gpu_power_draw"
+	GpuRequest               = "gpu_request"
+	GpuLimit                 = "gpu_limit"
+	GpuTotal                 = "gpu_total"
+	GpuUniqueID              = "UUID"
+	GpuTensorCoreUtilization = "gpu_tensor_core_utilization"
 
 	NeuronCoreUtilization                       = "neuroncore_utilization"
 	NeuronCoreMemoryUtilizationTotal            = "neuroncore_memory_usage_total"

plugins/processors/gpuattributes/internal/metricFilters/gpumetricfilters.go

@@ -30,7 +30,7 @@ var ContainerGpuLabelFilter = map[string]map[string]interface{}{
 	containerinsightscommon.FullPodNameKey:   nil,
 	containerinsightscommon.PodNameKey:       nil,
 	containerinsightscommon.TypeService:      nil,
-	containerinsightscommon.GpuUniqueId:      nil,
+	containerinsightscommon.GpuUniqueID:      nil,
 	containerinsightscommon.ContainerNamekey: nil,
 	containerinsightscommon.InstanceTypeKey:  nil,
 	containerinsightscommon.VersionKey:       nil,
@@ -57,7 +57,7 @@ var PodGpuLabelFilter = map[string]map[string]interface{}{
 	containerinsightscommon.FullPodNameKey:  nil,
 	containerinsightscommon.PodNameKey:      nil,
 	containerinsightscommon.TypeService:     nil,
-	containerinsightscommon.GpuUniqueId:     nil,
+	containerinsightscommon.GpuUniqueID:     nil,
 	containerinsightscommon.InstanceTypeKey: nil,
 	containerinsightscommon.VersionKey:      nil,
 	containerinsightscommon.SourcesKey:      nil,

translator/tocwconfig/sampleConfig/emf_and_kubernetes_with_gpu_config.yaml

@@ -185,6 +185,8 @@ exporters:
                 - node_gpu_limit
                 - node_gpu_usage_total
                 - node_gpu_reserved_capacity
+                - node_gpu_unreserved_capacity
+                - node_gpu_available_capacity
             - dimensions:
                 - - ClusterName
                   - InstanceId
@@ -346,6 +348,7 @@ exporters:
                 - container_gpu_memory_used
                 - container_gpu_power_draw
                 - container_gpu_temperature
+                - container_gpu_tensor_core_utilization
             - dimensions:
                 - - ClusterName
                 - - ClusterName
@@ -372,6 +375,7 @@ exporters:
                 - pod_gpu_memory_used
                 - pod_gpu_power_draw
                 - pod_gpu_temperature
+                - pod_gpu_tensor_core_utilization
             - dimensions:
                 - - ClusterName
                 - - ClusterName
@@ -389,6 +393,7 @@ exporters:
                 - node_gpu_memory_used
                 - node_gpu_power_draw
                 - node_gpu_temperature
+                - node_gpu_tensor_core_utilization
             - dimensions:
                 - - ClusterName
                 - - ClusterName
@@ -1226,6 +1231,48 @@ processors:
                   new_label: Type
                   new_value: NodeGPU
               submatch_case: ""
+            - action: insert
+              aggregation_type: ""
+              include: DCGM_FI_PROF_PIPE_TENSOR_ACTIVE
+              match_type: ""
+              new_name: container_gpu_tensor_core_utilization
+              operations:
+                - action: add_label
+                  aggregation_type: ""
+                  experimental_scale: 0
+                  label: ""
+                  label_value: ""
+                  new_label: Type
+                  new_value: ContainerGPU
+              submatch_case: ""
+            - action: insert
+              aggregation_type: ""
+              include: DCGM_FI_PROF_PIPE_TENSOR_ACTIVE
+              match_type: ""
+              new_name: pod_gpu_tensor_core_utilization
+              operations:
+                - action: add_label
+                  aggregation_type: ""
+                  experimental_scale: 0
+                  label: ""
+                  label_value: ""
+                  new_label: Type
+                  new_value: PodGPU
+              submatch_case: ""
+            - action: insert
+              aggregation_type: ""
+              include: DCGM_FI_PROF_PIPE_TENSOR_ACTIVE
+              match_type: ""
+              new_name: node_gpu_tensor_core_utilization
+              operations:
+                - action: add_label
+                  aggregation_type: ""
+                  experimental_scale: 0
+                  label: ""
+                  label_value: ""
+                  new_label: Type
+                  new_value: NodeGPU
+              submatch_case: ""
             - action: update
               aggregation_type: ""
               include: execution_status_total

translator/translate/otel/exporter/awsemf/kubernetes.go

@@ -154,7 +154,7 @@ func getNodeMetricDeclarations(conf *confmap.Conf) []*awsemfexporter.MetricDecla
 		"node_status_capacity_pods", "node_status_allocatable_pods",
 	}
 	if awscontainerinsight.AcceleratedComputeMetricsEnabled(conf) {
-		nodeMetrics = append(nodeMetrics, "node_gpu_limit", "node_gpu_usage_total", "node_gpu_reserved_capacity")
+		nodeMetrics = append(nodeMetrics, "node_gpu_limit", "node_gpu_usage_total", "node_gpu_reserved_capacity", "node_gpu_unreserved_capacity", "node_gpu_available_capacity")
 	}
 	if enhancedContainerInsightsEnabled {
 		return []*awsemfexporter.MetricDeclaration{
@@ -491,6 +491,7 @@ func getGPUMetricDeclarations(conf *confmap.Conf) []*awsemfexporter.MetricDeclar
 					"container_gpu_memory_used",
 					"container_gpu_power_draw",
 					"container_gpu_temperature",
+					"container_gpu_tensor_core_utilization",
 				},
 			},
 			{
@@ -502,6 +503,7 @@ func getGPUMetricDeclarations(conf *confmap.Conf) []*awsemfexporter.MetricDeclar
 					"pod_gpu_memory_used",
 					"pod_gpu_power_draw",
 					"pod_gpu_temperature",
+					"pod_gpu_tensor_core_utilization",
 				},
 			},
 			{
@@ -513,6 +515,7 @@ func getGPUMetricDeclarations(conf *confmap.Conf) []*awsemfexporter.MetricDeclar
 					"node_gpu_memory_used",
 					"node_gpu_power_draw",
 					"node_gpu_temperature",
+					"node_gpu_tensor_core_utilization",
 				},
 			},
 		}...)

translator/translate/otel/exporter/awsemf/translator_test.go

@@ -320,7 +320,7 @@ func TestTranslator(t *testing.T) {
 							"node_cpu_usage_total", "node_cpu_limit", "node_memory_working_set", "node_memory_limit",
 							"node_status_condition_ready", "node_status_condition_disk_pressure", "node_status_condition_memory_pressure",
 							"node_status_condition_pid_pressure", "node_status_condition_network_unavailable", "node_status_condition_unknown",
-							"node_status_capacity_pods", "node_status_allocatable_pods", "node_gpu_limit", "node_gpu_usage_total", "node_gpu_reserved_capacity"},
+							"node_status_capacity_pods", "node_status_allocatable_pods", "node_gpu_limit", "node_gpu_usage_total", "node_gpu_reserved_capacity", "node_gpu_unreserved_capacity", "node_gpu_available_capacity"},
 					},
 					{
 						Dimensions: [][]string{
@@ -405,19 +405,19 @@ func TestTranslator(t *testing.T) {
 					{
 						Dimensions: [][]string{{"ClusterName"}, {"ClusterName", "Namespace", "PodName", "ContainerName"}, {"ClusterName", "Namespace", "PodName", "FullPodName", "ContainerName"}, {"ClusterName", "Namespace", "PodName", "FullPodName", "ContainerName", "GpuDevice"}},
 						MetricNameSelectors: []string{
-							"container_gpu_utilization", "container_gpu_memory_utilization", "container_gpu_memory_total", "container_gpu_memory_used", "container_gpu_power_draw", "container_gpu_temperature",
+							"container_gpu_utilization", "container_gpu_memory_utilization", "container_gpu_memory_total", "container_gpu_memory_used", "container_gpu_power_draw", "container_gpu_temperature", "container_gpu_tensor_core_utilization",
 						},
 					},
 					{
 						Dimensions: [][]string{{"ClusterName"}, {"ClusterName", "Namespace"}, {"ClusterName", "Namespace", "Service"}, {"ClusterName", "Namespace", "PodName"}, {"ClusterName", "Namespace", "PodName", "FullPodName"}, {"ClusterName", "Namespace", "PodName", "FullPodName", "GpuDevice"}},
 						MetricNameSelectors: []string{
-							"pod_gpu_utilization", "pod_gpu_memory_utilization", "pod_gpu_memory_total", "pod_gpu_memory_used", "pod_gpu_power_draw", "pod_gpu_temperature",
+							"pod_gpu_utilization", "pod_gpu_memory_utilization", "pod_gpu_memory_total", "pod_gpu_memory_used", "pod_gpu_power_draw", "pod_gpu_temperature", "pod_gpu_tensor_core_utilization",
 						},
 					},
 					{
 						Dimensions: [][]string{{"ClusterName"}, {"ClusterName", "NodeName", "InstanceId"}, {"ClusterName", "NodeName", "InstanceId", "InstanceType", "GpuDevice"}},
 						MetricNameSelectors: []string{
-							"node_gpu_utilization", "node_gpu_memory_utilization", "node_gpu_memory_total", "node_gpu_memory_used", "node_gpu_power_draw", "node_gpu_temperature",
+							"node_gpu_utilization", "node_gpu_memory_utilization", "node_gpu_memory_total", "node_gpu_memory_used", "node_gpu_power_draw", "node_gpu_temperature", "node_gpu_tensor_core_utilization",
 						},
 					},
 					{

translator/translate/otel/processor/metricstransformprocessor/translator.go

@@ -30,12 +30,13 @@ var metricDuplicateTypes = []string{
 }
 
 var renameMapForDcgm = map[string]string{
-	"DCGM_FI_DEV_GPU_UTIL":        containerinsightscommon.GpuUtilization,
-	"DCGM_FI_DEV_FB_USED_PERCENT": containerinsightscommon.GpuMemUtilization,
-	"DCGM_FI_DEV_FB_USED":         containerinsightscommon.GpuMemUsed,
-	"DCGM_FI_DEV_FB_TOTAL":        containerinsightscommon.GpuMemTotal,
-	"DCGM_FI_DEV_GPU_TEMP":        containerinsightscommon.GpuTemperature,
-	"DCGM_FI_DEV_POWER_USAGE":     containerinsightscommon.GpuPowerDraw,
+	"DCGM_FI_DEV_GPU_UTIL":            containerinsightscommon.GpuUtilization,
+	"DCGM_FI_DEV_FB_USED_PERCENT":     containerinsightscommon.GpuMemUtilization,
+	"DCGM_FI_DEV_FB_USED":             containerinsightscommon.GpuMemUsed,
+	"DCGM_FI_DEV_FB_TOTAL":            containerinsightscommon.GpuMemTotal,
+	"DCGM_FI_DEV_GPU_TEMP":            containerinsightscommon.GpuTemperature,
+	"DCGM_FI_DEV_POWER_USAGE":         containerinsightscommon.GpuPowerDraw,
+	"DCGM_FI_PROF_PIPE_TENSOR_ACTIVE": containerinsightscommon.GpuTensorCoreUtilization,
 }
 
 var renameMapForNeuronMonitor = map[string]string{