kubernetes · k8s-ci-robot · Mar 14, 2026 · Mar 13, 2026 · justinsb · Mar 14, 2026
diff --git a/...s/ai-conformance/validators/observability/accelerator_metrics/accelerator_metrics_test.go b/...s/ai-conformance/validators/observability/accelerator_metrics/accelerator_metrics_test.go
@@ -0,0 +1,131 @@
+/*
+Copyright The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package ai_inference
+
+import (
+	"fmt"
+	"strings"
+	"testing"
+	"time"
+
+	"k8s.io/kops/tests/e2e/scenarios/ai-conformance/validators"
+)
+
+// TestObservability_AcceleratorMetrics corresponds to the observability/accelerator_metrics conformance requirement.
+func TestObservability_AcceleratorMetrics(t *testing.T) {
+	// Description:
+	// For supported accelerator types, the platform must allow for the installation and successful operation of at least one accelerator metrics solution
+	// that exposes fine-grained performance metrics via a standardized, machine-readable metrics endpoint.
+	// This must include a core set of metrics for per-accelerator utilization and memory usage.
+	// Additionally, other relevant metrics such as temperature, power draw, and interconnect bandwidth should be exposed
+	// if the underlying hardware or virtualization layer makes them available.
+	// The list of metrics should align with emerging standards, such as OpenTelemetry metrics, to ensure interoperability.
+	// The platform may provide a managed solution, but this is not required for conformance."
+
+	h := validators.NewValidatorHarness(t)
+
+	h.Logf("# Observability: Accelerator Metrics")
+
+	h.Run("nvidia-metrics", func(h *validators.ValidatorHarness) {
+		h.Logf("## Verify NVIDIA Metrics")
+
+		h.ShellExec("kubectl get service -n gpu-operator")
+
+		ns := h.TestNamespace()
+
+		requiredMetrics := []string{
+			"DCGM_FI_DEV_GPU_TEMP",
+			"DCGM_FI_DEV_POWER_USAGE",
+			"DCGM_FI_DEV_GPU_UTIL",
+			"DCGM_FI_DEV_FB_USED",
+		}
+
+		var metricClasses map[string]bool
+		var metricsOutput string
+
+		// Retry scraping metrics, as the DCGM exporter may not have completed its first collection cycle yet.
+		const maxAttempts = 5
+		for attempt := 1; attempt <= maxAttempts; attempt++ {
+			podName := fmt.Sprintf("scrape-accelerator-metrics-%d", attempt)
+
+			h.ShellExec(fmt.Sprintf(
+				"kubectl run %s -n %s --image=registry.k8s.io/e2e-test-images/agnhost:2.39 --restart=Never --command -- curl -sS http://nvidia-dcgm-exporter.gpu-operator.svc.cluster.local:9400/metrics",
+				podName, ns,
+			))
+			//h.ShellExec(fmt.Sprintf("kubectl wait -n %s pod/%s --for=condition=Ready --timeout=60s", ns, jobName))
+			h.ShellExec(fmt.Sprintf("kubectl wait -n %s pod/%s --for=jsonpath='{.status.phase}'=Succeeded --timeout=120s", ns, podName))
+
+			logs := h.ShellExec(fmt.Sprintf("kubectl logs -n %s %s", ns, podName))
+			metricsOutput = logs.Stdout()
+
+			metricClasses = make(map[string]bool)
+			for _, line := range strings.Split(metricsOutput, "\n") {
+				line = strings.TrimSpace(line)
+				// Ignore comment lines
+				if strings.HasPrefix(line, "#") {
+					continue
+				}
+				fields := strings.Fields(line)
+				// Ignore lines that don't have at least a metric name and value
+				if len(fields) < 2 {
+					continue
+				}
+				metric := fields[0]
+
+				// Extract out the metric class, ignoring any labels. For example, from "DCGM_FI_DEV_GPU_TEMP{gpu=\"0\"}" we want "DCGM_FI_DEV_GPU_TEMP".
+				metricClass := metric
+				if prefix, _, ok := strings.Cut(metric, "{"); ok {
+					metricClass = prefix
+				}
+
+				// Record the metric class as found
+				metricClasses[metricClass] = true
+			}
+
+			allFound := true
+			for _, m := range requiredMetrics {
+				if !metricClasses[m] {
+					allFound = false
+					break
+				}
+			}
+			if allFound {
+				h.Logf("All required metrics found on attempt %d", attempt)
+				break
+			}
+
+			if attempt < maxAttempts {
+				h.Logf("Attempt %d: not all required metrics found, retrying in 10s...", attempt)
+				time.Sleep(10 * time.Second)
+			}
+		}
+
+		h.Logf("Received metrics:\n%s", metricsOutput)
+
+		for _, m := range requiredMetrics {
+			if !metricClasses[m] {
+				h.Errorf("Did not find expected metric: %s", m)
+			} else {
+				h.Logf("Found expected metric: %s", m)
+			}
+		}
+	})
+
+	if h.AllPassed() {
+		h.RecordConformance("observability/accelerator_metrics")
+	}
+}