|
| 1 | +/* |
| 2 | +Copyright The Kubernetes Authors. |
| 3 | +
|
| 4 | +Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | +you may not use this file except in compliance with the License. |
| 6 | +You may obtain a copy of the License at |
| 7 | +
|
| 8 | + http://www.apache.org/licenses/LICENSE-2.0 |
| 9 | +
|
| 10 | +Unless required by applicable law or agreed to in writing, software |
| 11 | +distributed under the License is distributed on an "AS IS" BASIS, |
| 12 | +WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | +See the License for the specific language governing permissions and |
| 14 | +limitations under the License. |
| 15 | +*/ |
| 16 | + |
| 17 | +package ai_inference |
| 18 | + |
| 19 | +import ( |
| 20 | + "fmt" |
| 21 | + "strings" |
| 22 | + "testing" |
| 23 | + |
| 24 | + "k8s.io/kops/tests/e2e/scenarios/ai-conformance/validators" |
| 25 | +) |
| 26 | + |
| 27 | +// TestObservability_AcceleratorMetrics corresponds to the observability/accelerator_metrics conformance requirement. |
| 28 | +func TestObservability_AcceleratorMetrics(t *testing.T) { |
| 29 | + // Description: |
| 30 | + // For supported accelerator types, the platform must allow for the installation and successful operation of at least one accelerator metrics solution |
| 31 | + // that exposes fine-grained performance metrics via a standardized, machine-readable metrics endpoint. |
| 32 | + // This must include a core set of metrics for per-accelerator utilization and memory usage. |
| 33 | + // Additionally, other relevant metrics such as temperature, power draw, and interconnect bandwidth should be exposed |
| 34 | + // if the underlying hardware or virtualization layer makes them available. |
| 35 | + // The list of metrics should align with emerging standards, such as OpenTelemetry metrics, to ensure interoperability. |
| 36 | + // The platform may provide a managed solution, but this is not required for conformance." |
| 37 | + |
| 38 | + h := validators.NewValidatorHarness(t) |
| 39 | + |
| 40 | + h.Logf("# Observability: Accelerator Metrics") |
| 41 | + |
| 42 | + h.Run("nvidia-metrics", func(h *validators.ValidatorHarness) { |
| 43 | + h.Logf("## Verify NVIDIA Metrics") |
| 44 | + |
| 45 | + h.ShellExec("kubectl get service -n gpu-operator") |
| 46 | + |
| 47 | + ns := h.TestNamespace() |
| 48 | + |
| 49 | + objects := h.ApplyManifest("testdata/scrape-metrics.yaml", ns) |
| 50 | + for _, obj := range objects { |
| 51 | + obj.KubectlWait() |
| 52 | + } |
| 53 | + |
| 54 | + logs := h.ShellExec(fmt.Sprintf("kubectl logs -n %s -l=job-name=scrape-metrics", ns)) |
| 55 | + |
| 56 | + t.Logf("Received metrics:\n%s", string(logs.Stdout())) |
| 57 | + if !strings.Contains(string(logs.Stdout()), "DCGM_FI_DEV_GPU_UTIL") { |
| 58 | + h.Fatalf("Did not find expected GPU utilization metric in output: %s", logs.Stdout()) |
| 59 | + } else { |
| 60 | + h.Success("Found expected GPU utilization metric in output, indicating that accelerator metrics are being exposed correctly.") |
| 61 | + } |
| 62 | + |
| 63 | + }) |
| 64 | + |
| 65 | + if h.AllPassed() { |
| 66 | + h.RecordConformance("observability/accelerator_metrics") |
| 67 | + } |
| 68 | +} |
0 commit comments