Make image steps (with gpu driver)

Make a user's own virtual machine image that can be used to create multiple instances of a virtual machine. To create a virtual machine template follow the steps below

Step 1: Installing KVM in Linux

1. Install kvm and other software on linux host

sudo apt install qemu-kvm libvirt-clients libvirt-daemon-system bridge-utils virt-manager ovmf cpu-checker vim expect -y

2. Check virtualization support

After installing kvm and other software, you need to make sure the current machine supports kvm, use the following command to check virtualization support

egrep -c '(vmx|svm)' /proc/cpuinfo

Any non-zero result means virtualization is supported. From the output below, we have confirmed that our server is working properly.

Then check the system support for kvm virtualization, using the following command

sudo kvm-ok

image_1

If you have the output of the above picture, you can proceed, if not please check the host bios settings

3. VFIO Preparation

Set kernel module blacklist

sudo vim /etc/modprobe.d/blacklist.conf
#Add these at the bottom of the file：
blacklist snd_hda_intel
blacklist amd76x_edac
blacklist vga16fb
blacklist nouveau
blacklist rivafb
blacklist nvidiafb
blacklist rivatv

Set gpu passthrough

#search gpu ID
lspci -nnv|grep -i NVIDIA
#Record the graphics ID as shown below ,like 10de:2204 10de:1aef

image_2

#Modify kernel file
sudo vim /etc/default/grub
#Add in GRUB_CMDLINE_LINUX_DEFAULT field (if AMD platform, intel_iommu=on to amd_iommu=on)
#”quiet splash intel_iommu=on kvm.ignore_msrs=1 vfio-pci.ids=graphics ID <--Separated by commas“

image_3

#update kernel
sudo update-grub

Then restart the machine and check the graphics card usage after the machine starts

#4f:00.0 as an example
lspci -vv -s 4f:00 |grep driver

If vfio-pci is shown after driver, it is correct

image_4

4. Checking libvirt running

Check that the libvirt service is running, if it is not, you need to start it manually

sudo systemctl status libvirtd

image_5

You can enable it to start if not running:

sudo systemctl enable --now libvirtd

Step 2: Create a KVM Virtual Image

Before we get to creating a template, we need to, first of all, have an installation instance. On the command-line, we are going to create a 50G Ubuntu KVM image using the qemu-img command as shown.

sudo qemu-img create -f qcow2 /data/ubuntu.qcow2 50G

Then create the ubuntu virtual machine with the virt-install command as shown below. Note that the iso file after the "--cdrom" is the installation image of the OS you want to install, you need to download it yourself

virt-install --virt-type kvm --name ubuntu --memory 4096 \
--vcpus 4 --cpu=host-passthrough \
--cdrom=/data/ubuntu_install_dvd_base.iso \
--os-variant=auto \
--disk /data/ubuntu.qcow2,bus=virtio \
--network network=default \
--graphics vnc,listen=0.0.0.0 --noautoconsole \
--host-device 4f:00.0 \
--host-device 4f:00.1

image_6

This launches the virtual machine instance. You can confirm this by heading over to the vnc viewer and opening the console window as shown. What you can see is the default welcome page for the installer. Be sure to complete the installation to the very end.

image_7

Step 3: Create KVM Virtual Machine Template Image

Once the installation is complete, log into the VM and update all the system packages.

sudo apt update

Install the prerequisite packages that you feel are essential to get started with.(Be sure to install qemu-guest-agent and nvidia gpu driver)

sudo apt install qemu-guest-agent
sudo systemctl enable qemu-guest-agent --now
#NVIDIA graphics card drivers need to go to the official website to download the corresponding version
#Or use apt install to install nvidia drivers

Once you are done, be sure to power off your virtual machine and clean up the VM template image as shown.

sudo virt-sysprep --enable bash-history,tmp-files,machine-id -d ubuntu

image_8

Lastly, invoke the command shown to undefine the VM domain. And he template image is now ready for cloning and deployment.