|
| 1 | +/* |
| 2 | + * Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved. |
| 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 e2e |
| 18 | + |
| 19 | +import ( |
| 20 | + "context" |
| 21 | + |
| 22 | + . "github.com/onsi/ginkgo/v2" |
| 23 | + . "github.com/onsi/gomega" |
| 24 | +) |
| 25 | + |
| 26 | +// Integration tests for Docker runtime |
| 27 | +var _ = Describe("docker", func() { |
| 28 | + // GPUs are accessible in a container: Running nvidia-smi -L inside the |
| 29 | + // container shows the same output inside the container as outside the |
| 30 | + // container. This means that the following commands must all produce |
| 31 | + // the same output |
| 32 | + When("running nvidia-smi -L", Ordered, func() { |
| 33 | + var hostOutput string |
| 34 | + |
| 35 | + BeforeAll(func(ctx context.Context) { |
| 36 | + _, err := runScript("docker pull ubuntu") |
| 37 | + Expect(err).ToNot(HaveOccurred()) |
| 38 | + |
| 39 | + hostOutput, err = runScript("nvidia-smi -L") |
| 40 | + Expect(err).ToNot(HaveOccurred()) |
| 41 | + }) |
| 42 | + |
| 43 | + It("should support NVIDIA_VISIBLE_DEVICES", func(ctx context.Context) { |
| 44 | + containerOutput, err := runScript("docker run --rm -i --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=all ubuntu nvidia-smi -L") |
| 45 | + Expect(err).ToNot(HaveOccurred()) |
| 46 | + Expect(containerOutput).To(Equal(hostOutput)) |
| 47 | + }) |
| 48 | + |
| 49 | + It("should support automatic CDI spec generation", func(ctx context.Context) { |
| 50 | + containerOutput, err := runScript("docker run --rm -i --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=runtime.nvidia.com/gpu=all ubuntu nvidia-smi -L") |
| 51 | + Expect(err).ToNot(HaveOccurred()) |
| 52 | + Expect(containerOutput).To(Equal(hostOutput)) |
| 53 | + }) |
| 54 | + |
| 55 | + It("should support the --gpus flag using the nvidia-container-runtime", func(ctx context.Context) { |
| 56 | + containerOutput, err := runScript("docker run --rm -i --runtime=nvidia --gpus all ubuntu nvidia-smi -L") |
| 57 | + Expect(err).ToNot(HaveOccurred()) |
| 58 | + Expect(containerOutput).To(Equal(hostOutput)) |
| 59 | + }) |
| 60 | + |
| 61 | + It("should support the --gpus flag using the nvidia-container-runtime-hook", func(ctx context.Context) { |
| 62 | + containerOutput, err := runScript("docker run --rm -i --gpus all ubuntu nvidia-smi -L") |
| 63 | + Expect(err).ToNot(HaveOccurred()) |
| 64 | + Expect(containerOutput).To(Equal(hostOutput)) |
| 65 | + }) |
| 66 | + }) |
| 67 | + |
| 68 | + // A vectorAdd sample runs in a container with access to all GPUs. |
| 69 | + // The following should all produce the same result. |
| 70 | + When("Running the cuda-vectorAdd sample", Ordered, func() { |
| 71 | + BeforeAll(func(ctx context.Context) { |
| 72 | + _, err := runScript("docker pull nvcr.io/nvidia/k8s/cuda-sample:vectoradd-cuda12.5.0") |
| 73 | + Expect(err).ToNot(HaveOccurred()) |
| 74 | + }) |
| 75 | + |
| 76 | + var referenceOutput string |
| 77 | + |
| 78 | + It("should support NVIDIA_VISIBLE_DEVICES", func(ctx context.Context) { |
| 79 | + var err error |
| 80 | + referenceOutput, err = runScript("docker run --rm -i --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=all nvcr.io/nvidia/k8s/cuda-sample:vectoradd-cuda12.5.0") |
| 81 | + Expect(err).ToNot(HaveOccurred()) |
| 82 | + |
| 83 | + Expect(referenceOutput).To(ContainSubstring("Test PASSED")) |
| 84 | + }) |
| 85 | + |
| 86 | + It("should support automatic CDI spec generation", func(ctx context.Context) { |
| 87 | + out2, err := runScript("docker run --rm -i --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=runtime.nvidia.com/gpu=all nvcr.io/nvidia/k8s/cuda-sample:vectoradd-cuda12.5.0") |
| 88 | + Expect(err).ToNot(HaveOccurred()) |
| 89 | + Expect(referenceOutput).To(Equal(out2)) |
| 90 | + }) |
| 91 | + |
| 92 | + It("should support the --gpus flag using the nvidia-container-runtime", func(ctx context.Context) { |
| 93 | + out3, err := runScript("docker run --rm -i --runtime=nvidia --gpus all nvcr.io/nvidia/k8s/cuda-sample:vectoradd-cuda12.5.0") |
| 94 | + Expect(err).ToNot(HaveOccurred()) |
| 95 | + Expect(referenceOutput).To(Equal(out3)) |
| 96 | + }) |
| 97 | + |
| 98 | + It("should support the --gpus flag using the nvidia-container-runtime-hook", func(ctx context.Context) { |
| 99 | + out4, err := runScript("docker run --rm -i --gpus all nvcr.io/nvidia/k8s/cuda-sample:vectoradd-cuda12.5.0") |
| 100 | + Expect(err).ToNot(HaveOccurred()) |
| 101 | + Expect(referenceOutput).To(Equal(out4)) |
| 102 | + }) |
| 103 | + }) |
| 104 | + |
| 105 | + // A deviceQuery sample runs in a container with access to all GPUs |
| 106 | + // The following should all produce the same result. |
| 107 | + When("Running the cuda-deviceQuery sample", Ordered, func() { |
| 108 | + BeforeAll(func(ctx context.Context) { |
| 109 | + _, err := runScript("docker pull nvcr.io/nvidia/k8s/cuda-sample:devicequery-cuda12.5.0") |
| 110 | + Expect(err).ToNot(HaveOccurred()) |
| 111 | + }) |
| 112 | + |
| 113 | + var referenceOutput string |
| 114 | + |
| 115 | + It("should support NVIDIA_VISIBLE_DEVICES", func(ctx context.Context) { |
| 116 | + var err error |
| 117 | + referenceOutput, err = runScript("docker run --rm -i --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=all nvcr.io/nvidia/k8s/cuda-sample:devicequery-cuda12.5.0") |
| 118 | + Expect(err).ToNot(HaveOccurred()) |
| 119 | + |
| 120 | + Expect(referenceOutput).To(ContainSubstring("Result = PASS")) |
| 121 | + }) |
| 122 | + |
| 123 | + It("should support automatic CDI spec generation", func(ctx context.Context) { |
| 124 | + out2, err := runScript("docker run --rm -i --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=runtime.nvidia.com/gpu=all nvcr.io/nvidia/k8s/cuda-sample:devicequery-cuda12.5.0") |
| 125 | + Expect(err).ToNot(HaveOccurred()) |
| 126 | + Expect(referenceOutput).To(Equal(out2)) |
| 127 | + }) |
| 128 | + |
| 129 | + It("should support the --gpus flag using the nvidia-container-runtime", func(ctx context.Context) { |
| 130 | + out3, err := runScript("docker run --rm -i --runtime=nvidia --gpus all nvcr.io/nvidia/k8s/cuda-sample:devicequery-cuda12.5.0") |
| 131 | + Expect(err).ToNot(HaveOccurred()) |
| 132 | + Expect(referenceOutput).To(Equal(out3)) |
| 133 | + }) |
| 134 | + |
| 135 | + It("should support the --gpus flag using the nvidia-container-runtime-hook", func(ctx context.Context) { |
| 136 | + out4, err := runScript("docker run --rm -i --gpus all nvcr.io/nvidia/k8s/cuda-sample:devicequery-cuda12.5.0") |
| 137 | + Expect(err).ToNot(HaveOccurred()) |
| 138 | + Expect(referenceOutput).To(Equal(out4)) |
| 139 | + }) |
| 140 | + }) |
| 141 | +}) |
0 commit comments