This guide will teach you how to create and manage a VM in KVM using virt-install, virsh, and virt-viewer, which all make use of libvirt. For a beginner's guide and introduction to libvirt and virtualisation concepts, and why you would even want to use libvirt instead of, say, Virtualbox, see: https://jamesnorth.net/post/qemu-guide
You can absolutely setup and manage QEMU domains purely with virsh by editing the XML files manually and connecting to the console with virsh as opposed to using virt-viewer. I don't recommend this for the faint of heart. We'll be using virt-install
and virt-viewer
for the heavy lifting.
Table of Contents
#Create Your VM
Install Prerequisites
You need a few packages before you can start working with libvirt
. Note that the spice
and virt-viewer
packages are optional if you only intend to work with your guests directly through the console.
On Arch:
# pacman -S libvirt qemu-desktop virt-install iptables-nft dnsmasq virt-viewer spice spice-gtk # systemctl enable --now libvirtd
Then reboot your machine to load the kernel modules that these packages install/setup.
Obtain the ISO
Download the ISO for the VM you want to install. My choice is Gentoo: https://www.gentoo.org/downloads/
Create the VM
Make sure the default network is active with:
# virsh net-start default
Use virt-install
to create the VM image and XML file:
# virt-install -n yuki --connect qemu:///system --description "gentoo install test" --osinfo=gentoo --memory=16384--vcpus=4 --disk path=/var/lib/libvirt/images/yuki.img,bus=virtio,size=40 --graphics spice --cdrom /var/install-amd64-minimal.iso --network network=default
This command will create a Domain (libvirt's terminology for "VM Guest") with the name "yuki" and description "gentoo install test", connect to it via qemu:///system, and knows that it's installing a gentoo guest because of the osinfo parameter. It will have 16GB of RAM, 4 vcpus (CPU cores, essentially), create or use the image at /var/lib/libvirt/images/yuki.img if it already exists, using the virtio bus for QEMU, the image will be 40GB, it will use the iso at /var/install-amd64-minimal.iso to install the OS, and it will use the default network.
Not all of these options are necessary, and more options can be specified for greater control. You can find them in the virt-install (1) man page.
To find out what your OS's Short ID is, run osinfo-query os
. You can also filter the output to only the information useful to you: osinfo-query os | grep gentoo
This command will create the VM domain and launch the live installation media in virt-viewer (if not, you can try virt-viewer yuki
to connect to it). You can now run through the install process.
Re-inserting Installation Media
If you shut down the domain for any reason instead of installing the OS immediately, the installation media will no longer be inserted and you won't be able to boot the next time you start the domain.
To fix this, use virsh change-media
to insert the installation media. Before that, you will need to know the disks available on your domain, so use:
# virsh domblkinfo --all yuki
The target you are looking for is generally 'sda' or 'sdb'. Now you can try inserting the installation media:
# virsh change-media yuki sda --insert /var/installation-amd64-minimal.iso
You will need to alter the domain's xml configuration to boot from the cdrom (which is what sda is):
# virsh edit yuki
Insert <boot dev='cdrom'/>
below <boot dev='hd'/>
and save the file. This will attempt to boot from the 'hd' device first (which is vda, as you'll see further down the file), and if that fails (which it will, as it is currently empty), boot from the cdrom. This is useful because we want to boot from the disk after we install the OS, and this means we don't have to edit the file again.
You can now start your domain and connect to it:
# virsh start yuki && virt-viewer -w yuki
You can also determine the boot order using per-device boot elements, which are mutually exclusive with the domain-wide boot element I've just explained. More on that and other domain booting information (also see bootmenu) here: https://libvirt.org/formatdomain.html#operating-system-booting
Per-device boot elements are helpful if you have multiple disks.
Pro-Tip: You can access much of the documentation for libvirt offline at /usr/share/docs/libvirt/html/docs.html (the location may differ depending on your distribution's filesystem setup)
Manage Your VM
When you're done with your VM, you can power it off from within the VM guest, or you can turn it off using virsh:
Graceful Shutdown and More
virsh can coordinate with the domain to power off gracefully, which is recommended unless something is blocking shutdown:
# virsh shutdown yuki
Or graceful reboot:
# virsh reboot yuki
Or suspend it gracefully:
# virsh suspend yuki
Resume it:
# virsh resume yuki
Force Off
The guest may need to be hard powered off, which is the violent alternative to the previous section:
# virsh destroy yuki
Delete (Undefine) a Domain
If you want to completely delete a domain, all you need to do is get rid of the XML file that defines it, as that's all a domain really is. With virsh:
# virsh undefine yuki
More With Virsh
See the man page for virsh (1) or try virsh help
for a more manageable list of actionable commands as a good place to start.