一、软硬件信息
1.服务器厂家:
2.沐曦GPU型号：C550
3.操作系统内核版本：Linux mx-oam-151 5.15.0-58-generic #64~20.04.1-Ubuntu SMP Fri Jan 6 16:42:31 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux
4.是否开启CPU虚拟化：开启
二、具体问题1
GPU-Operator部署后使用的SDK如图，
创建的pod如下
apiVersion: v1
kind: Pod
metadata:
name: task-sample-pod-2
spec:
schedulerName: hami-scheduler
containers:
- name: ubuntu-task
image: docker.io/ubuntu:20.04
imagePullPolicy: Never
command: [
"bash",
"-c",
"cp -r /opt/maca/samples/0_Introduction/vectorAdd /home;
cd /home/vectorAdd;
mxcc -x maca vectorAdd.cpp -o vectorAdd --maca-path=/opt/maca;
./vectorAdd > log/vectoradd_exec_output.log;
tail -f /dev/null",
]
resources:
limits:
metax-tech.com/sgpu: 1 # requesting 1 GPU
metax-tech.com/vcore: 40 # requesting 60% compute of full GPU
metax-tech.com/vmemory: 4 # requesting 4 GiB device memory of full GPU
报错如下：
三、具体问题2
如果新创建一个 pod，会显示
Allocate failed due to rpc error: code = Unknown desc = set 0000:c2:00.0 model error write /proc/1/root/sys/bus/pci/devices/0000:c2:00.0/model: device or resource busy, which is unexpected
但mx-smi如下，并没有使用第二张卡
