GPU机器搭建k8s
大约 3 分钟
gpu机器搭建k8s
查看GPU
(base) root@ubuntu:~# lspci | grep -i nvidia
01:00.0 VGA compatible controller: NVIDIA Corporation Device 2230 (rev a1)
01:00.1 Audio device: NVIDIA Corporation Device 1aef (rev a1)
22:00.0 VGA compatible controller: NVIDIA Corporation Device 2230 (rev a1)
22:00.1 Audio device: NVIDIA Corporation Device 1aef (rev a1)
41:00.0 VGA compatible controller: NVIDIA Corporation Device 2230 (rev a1)
41:00.1 Audio device: NVIDIA Corporation Device 1aef (rev a1)
61:00.0 VGA compatible controller: NVIDIA Corporation Device 2230 (rev a1)
61:00.1 Audio device: NVIDIA Corporation Device 1aef (rev a1)
可以看到有四张nvidia显卡和nvidia声卡
查看是否已安装gpu nvidia驱动
(base) root@ubuntu:~# lsmod | grep -i nvidia
nvidia_uvm 1216512 16
61440 0
nvidia_modeset 1241088 1 nvidia_drm
nvidia 56471552 1128 nvidia_uvm,nvidia_modeset
drm_kms_helper 184320 4 ast,nvidia_drm
drm 495616 7 drm_kms_helper,drm_vram_helper,ast,nvidia,nvidia_drm,ttm
此时已安装nvidia驱动
如果未安装：参考链接进行安装
查看gpu型号、gpu驱动版本
(base) root@ubuntu:~# nvidia-smi
Wed May 29 03:28:30 2024
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.116.04 Driver Version: 525.116.04 CUDA Version: 12.0 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA RTX A6000 Off | 00000000:01:00.0 Off | Off |
| 30% 29C P8 12W / 300W | 8142MiB / 49140MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
| 1 NVIDIA RTX A6000 Off | 00000000:22:00.0 Off | Off |
| 30% 27C P8 9W / 300W | 27682MiB / 49140MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
| 2 NVIDIA RTX A6000 Off | 00000000:41:00.0 Off | Off |
| 30% 28C P8 12W / 300W | 28252MiB / 49140MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
| 3 NVIDIA RTX A6000 Off | 00000000:61:00.0 Off | Off |
| 30% 26C P8 12W / 300W | 2MiB / 49140MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 1228160 C python 772MiB |
| 0 N/A N/A 1228264 C /usr/bin/python 3576MiB |
| 0 N/A N/A 4009358 C python 1896MiB |
| 0 N/A N/A 4009359 C python 1896MiB |
| 1 N/A N/A 1084495 C python 27680MiB |
| 2 N/A N/A 3972345 C python 28250MiB |
+-----------------------------------------------------------------------------+
可以看到四张显卡，对应编号是：0、1、2、3，其中：0、1、2三张卡已经有python进程正在使用。
查看docker运行时
(base) root@ubuntu:~# docker info | grep -i 'Default Runtime'
WARNING: No swap limit support
Default Runtime: nvidia
docker默认的运行时是：runc，为了使用gpu资源，需要使用：nvidia作为docker运行时。
如果未安装docker，参考：链接进行安装
使用minikube安装k8s
下载minikube
curl -LO https://storage.googleapis.com/minikube/releases/latest/minikube-linux-amd64
sudo install minikube-linux-amd64 /usr/local/bin/minikube
启动minikube
minikube start --image-mirror-country='cn' --image-repository='registry.cn-hangzhou.aliyuncs.com/google_containers' --force --driver docker --container-runtime docker --gpus all
其中：
- --force，原因是使用root用户，报错信息：The "docker" driver should not be used with root privileges. If you wish to continue as root, use --force.
- --driver docker，采用docker作为容器运行时
- --image-repository=registry.cn-hangzhou.aliyuncs.com/google_containers，由于国内环境无法访问国外镜像，采用阿里云的镜像仓库
- --gpus，all指定所有显卡
查看当前k8s集群
(base) root@ubuntu:/home/k8s# kubectl get pods -A
NAMESPACE NAME READY STATUS RESTARTS AGE
kube-system coredns-7c445c467-d6zsw 1/1 Running 0 24s
kube-system etcd-minikube 1/1 Running 0 40s
kube-system kube-apiserver-minikube 1/1 Running 0 37s
kube-system kube-controller-manager-minikube 1/1 Running 0 39s
kube-system kube-proxy-lblvn 1/1 Running 0 25s
kube-system kube-scheduler-minikube 1/1 Running 0 37s
kube-system storage-provisioner 1/1 Running 1 (8s ago) 35s
开启nvidia插件
minikube addons enable nvidia-device-plugin
或者
手动安装插件
kubectl apply -f https://raw.githubusercontent.com/NVIDIA/k8s-device-plugin/v0.15.0/deployments/static/nvidia-device-plugin.yml
设置别名
alias kubectl="minikube kubectl --"
验证是否能否设别GPU资源
kubectl describe nodes
部署应用使用gpu
apiVersion: v1
kind: Pod
metadata:
name: gpu
spec:
containers:
- name: gpu-container
image: nvidia/cuda:12.2.2-devel-ubuntu22.04
command:
- "/bin/sh"
- "-c"
args:
- nvidia-smi && tail -f /dev/null
resources:
requests:
nvidia.com/gpu: 1
limits:
nvidia.com/gpu: 1
kubectl apply -f gpu-deploy.yaml
查看日志
kubectl logs gpu
(base) root@ubuntu:~# kubectl logs gpu
Fri May 31 07:19:23 2024
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.116.04 Driver Version: 525.116.04 CUDA Version: 12.2 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA RTX A6000 Off | 00000000:01:00.0 Off | Off |
| 30% 29C P8 12W / 300W | 3794MiB / 49140MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
+-----------------------------------------------------------------------------+
其他问题
如果无法使用本地镜像，使用以下
eval $(minikube docker-env)