feat(go.d/nvidia_smi): add "index" label to GPU charts (netdata#18833)

webfutureiorepo · Oct 21, 2024 · c22ad52 · c22ad52
1 parent f0297a5
commit c22ad52
Show file tree

Hide file tree

Showing 3 changed files with 9 additions and 6 deletions.
diff --git a/src/go/plugin/go.d/modules/nvidia_smi/charts.go b/src/go/plugin/go.d/modules/nvidia_smi/charts.go
@@ -4,6 +4,7 @@ package nvidia_smi
 
 import (
 	"fmt"
+	"strconv"
 	"strings"
 
 	"github.com/netdata/netdata/go/plugins/plugin/go.d/agent/module"
@@ -261,7 +262,7 @@ var (
 	}
 )
 
-func (nv *NvidiaSmi) addGPUXMLCharts(gpu gpuInfo) {
+func (nv *NvidiaSmi) addGpuCharts(gpu gpuInfo, index int) {
 	charts := gpuXMLCharts.Copy()
 
 	if !isValidValue(gpu.Utilization.GpuUtil) {
@@ -294,7 +295,7 @@ func (nv *NvidiaSmi) addGPUXMLCharts(gpu gpuInfo) {
 	for _, c := range *charts {
 		c.ID = fmt.Sprintf(c.ID, strings.ToLower(gpu.UUID))
 		c.Labels = []module.Label{
-			// csv output has no 'product_brand'
+			{Key: "index", Value: strconv.Itoa(index)},
 			{Key: "uuid", Value: gpu.UUID},
 			{Key: "product_name", Value: gpu.ProductName},
 		}

diff --git a/src/go/plugin/go.d/modules/nvidia_smi/collect.go b/src/go/plugin/go.d/modules/nvidia_smi/collect.go
@@ -38,7 +38,7 @@ func (nv *NvidiaSmi) collectGPUInfo(mx map[string]int64) error {
 	seenGPU := make(map[string]bool)
 	seenMIG := make(map[string]bool)
 
-	for _, gpu := range info.GPUs {
+	for i, gpu := range info.GPUs {
 		if !isValidValue(gpu.UUID) {
 			continue
 		}
@@ -49,7 +49,7 @@ func (nv *NvidiaSmi) collectGPUInfo(mx map[string]int64) error {
 
 		if !nv.gpus[px] {
 			nv.gpus[px] = true
-			nv.addGPUXMLCharts(gpu)
+			nv.addGpuCharts(gpu, i)
 		}
 
 		addMetric(mx, px+"pcie_bandwidth_usage_rx", gpu.PCI.RxUtil, 1024) // KB => bytes

diff --git a/src/go/plugin/go.d/modules/nvidia_smi/metadata.yaml b/src/go/plugin/go.d/modules/nvidia_smi/metadata.yaml
@@ -98,7 +98,9 @@ modules:
           description: These metrics refer to the GPU.
           labels:
             - name: uuid
-              description: GPU id (e.g. 00000000:00:04.0)
+              description: GPU uuid (e.g. GPU-27b94a00-ed54-5c24-b1fd-1054085de32a)
+            - name: index
+              description: GPU index (nvidia_smi typically orders GPUs by PCI bus ID)
             - name: product_name
               description: GPU product name (e.g. NVIDIA A100-SXM4-40GB)
           metrics:
@@ -211,7 +213,7 @@ modules:
           description: These metrics refer to the Multi-Instance GPU (MIG).
           labels:
             - name: uuid
-              description: GPU id (e.g. 00000000:00:04.0)
+              description: GPU uuid (e.g. GPU-27b94a00-ed54-5c24-b1fd-1054085de32a)
             - name: product_name
               description: GPU product name (e.g. NVIDIA A100-SXM4-40GB)
             - name: gpu_instance_id