1)下载vGPU驱动(注:该驱动为付费商业软件,可以申请免费试用三个月):
该驱动包包含如下驱动:
host驱动:NVIDIA-Linux-x86_64-535.216.01-vgpu-kvm.run
guest驱动:NVIDIA-Linux-x86_64-535.216.01-grid.run
2)在安装有物理GPU卡的服务器上安装vGPU驱动
[root@localhost ~]# lspci|grep -i nvidia
98:00.0 3D controller: NVIDIA Corporation GV100GL [Tesla V100 SXM2 16GB] (rev a1)
[root@localhost ~]#
[root@localhost ~]#sh ./NVIDIA-Linux-x86_64-535.216.01-vgpu-kvm.run
3)确认nvidia_vgpu_vfio内核模块已被加载
[root@localhost vGPU16]# lsmod|grep -i vfio
nvidia_vgpu_vfio 122880 0
vfio_pci_core 90112 1 nvidia_vgpu_vfio
mdev 20480 1 nvidia_vgpu_vfio
vfio_iommu_type1 57344 0
vfio 73728 3 vfio_pci_core,nvidia_vgpu_vfio,vfio_iommu_type1
iommufd 110592 1 vfio
kvm 1380352 2 nvidia_vgpu_vfio,kvm_intel
irqbypass 12288 3 vfio_pci_core,nvidia_vgpu_vfio,kvm
[root@localhost vGPU16]#
4)启动nvidia-vgpu-mgr服务,并设置开机启动.
[root@localhost vGPU16]# systemctl enable nvidia-vgpu-mgr.service
Created symlink /etc/systemd/system/multi-user.target.wants/nvidia-vgpu-mgr.service → /usr/lib/systemd/system/nvidia-vgpu-mgr.service.
[root@localhost vGPU16]# systemctl start nvidia-vgpu-mgr.service
5)重启物理机器
[root@localhost vGPU16]# reboot
6)确认内核mdev目录是否生成,如果该目录下的内容为空,则表示没生成。
[root@localhost ~]# ls /sys/class/mdev_bus/
0000:98:00.0
[root@localhost ~]#