Fix unit test

aws · Oct 8, 2024 · dfa6d78 · dfa6d78
1 parent c5e7cc7
commit dfa6d78
Show file tree

Hide file tree

Showing 6 changed files with 11 additions and 1 deletion.
diff --git a/e2e2/test/cases/nvidia/manifests/job-unit-test-single-node.yaml b/e2e2/test/cases/nvidia/manifests/job-unit-test-single-node.yaml
@@ -21,6 +21,7 @@ spec:
           limits:
             cpu: "4"
             memory: 4Gi
+            nvidia.com/gpu: {{.GpuPerNode}}
           requests:
             cpu: "1"
             memory: 1Gi

diff --git a/e2e2/test/images/nvidia/Dockerfile b/e2e2/test/images/nvidia/Dockerfile
@@ -41,7 +41,8 @@ RUN apt install -y \
   cmake \
   apt-utils \
   libhwloc-dev \
-  cuda-demo-suite-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION}
+  cuda-demo-suite-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+  datacenter-gpu-manager
 
 RUN mkdir -p /var/run/sshd \
   && sed -i 's/[ #]\(.*StrictHostKeyChecking \).*/ \1no/g' /etc/ssh/ssh_config \

diff --git a/e2e2/test/images/nvidia/gpu_unit_tests/tests/test_sysinfo.sh.data/g5.8xlarge/gpu_count.txt b/e2e2/test/images/nvidia/gpu_unit_tests/tests/test_sysinfo.sh.data/g5.8xlarge/gpu_count.txt
@@ -0,0 +1,2 @@
+name, index, pci.bus_id
+NVIDIA A10G, 0, 00000000:00:1E.0
diff --git a/e2e2/test/images/nvidia/gpu_unit_tests/tests/test_sysinfo.sh.data/g5.8xlarge/numa_topo.txt b/e2e2/test/images/nvidia/gpu_unit_tests/tests/test_sysinfo.sh.data/g5.8xlarge/numa_topo.txt
@@ -0,0 +1,2 @@
+/sys/devices/system/node/node0/cpulist:0-31
+/sys/devices/system/node/node0/distance:10
diff --git a/...nvidia/gpu_unit_tests/tests/test_sysinfo.sh.data/g5.8xlarge/nvidia_persistence_status.txt b/...nvidia/gpu_unit_tests/tests/test_sysinfo.sh.data/g5.8xlarge/nvidia_persistence_status.txt
@@ -0,0 +1,2 @@
+name, pci.bus_id, persistence_mode
+NVIDIA A10G, 00000000:00:1E.0, Enabled
diff --git a/...st/images/nvidia/gpu_unit_tests/tests/test_sysinfo.sh.data/g5.8xlarge/nvidia_smi_topo.txt b/...st/images/nvidia/gpu_unit_tests/tests/test_sysinfo.sh.data/g5.8xlarge/nvidia_smi_topo.txt
@@ -0,0 +1,2 @@
+	[4mGPU0	CPU Affinity	NUMA Affinity	GPU NUMA ID[0m
+GPU0	 X 	0-31	0		N/A