An In-Depth Guide to Implementing Kernel-based Virtual Machine (KVM) on Linux

KVM (Kernel-based Virtual Machine) turns a Linux server into a hypervisor, so one physical machine can run many isolated virtual machines. This guide sets up KVM with libvirt on a current Ubuntu LTS server, creates a VM, and covers the everyday jobs: networking, storage, snapshots and clean-up. It is for your own server, where you have root access and hardware virtualization; it cannot be done on shared hosting.

1. What KVM is and how it works

KVM is a set of Linux kernel modules: kvm plus kvm_intel or kvm_amd, depending on the processor. They let the kernel use the CPU's hardware virtualization features (Intel VT-x or AMD-V) to run guest operating systems directly on the processor.

Three pieces work together:

Because each VM is a Linux process, the host's scheduler, memory management, cgroups, firewall and SELinux or AppArmor all apply to it. KVM is used by most large cloud platforms and by Proxmox VE, and it is the virtualization behind Domain India VPS plans.

2. Requirements

• A 64-bit x86 server with hardware virtualization enabled in the BIOS or UEFI (Intel VT-x or AMD-V).
• A current Linux distribution. This guide uses Ubuntu 24.04 LTS. On AlmaLinux or Rocky Linux the packages are qemu-kvm, libvirt and virt-install, installed with dnf.
• Enough RAM and disk for the host plus every VM you plan to run. Leave at least 1 to 2 GB of RAM for the host itself.
• Root or sudo access. On shared hosting, you cannot load kernel modules or run a hypervisor.

Check the CPU:

grep -Ec '(vmx|svm)' /proc/cpuinfo    # more than 0 means the CPU supports it
sudo apt install -y cpu-checker
kvm-ok                                # should say "KVM acceleration can be used"

If kvm-ok says virtualization is disabled, turn it on in the firmware. If you are inside a VM already, the host must offer nested virtualization.

3. Install KVM and libvirt

sudo apt update
sudo apt install -y qemu-kvm libvirt-daemon-system libvirt-clients virtinst
sudo usermod -aG libvirt,kvm "$USER"

Log out and back in so the group change applies. Then verify:

lsmod | grep kvm                  # kvm_intel or kvm_amd plus kvm
sudo systemctl is-active libvirtd # active
sudo virsh list --all             # empty list, no error

virtinst provides virt-install. You no longer need bridge-utils: bridges are created with netplan or the ip command, and brctl is deprecated.

4. Create your first VM

The quickest reliable way is to boot a cloud image with cloud-init, which gives you a ready server with your SSH key in a minute. Download the current Ubuntu LTS cloud image once:

cd /var/lib/libvirt/images
sudo wget https://cloud-images.ubuntu.com/noble/current/noble-server-cloudimg-amd64.img

Create a VM with 2 GB of RAM, 2 vCPUs and a 20 GB disk layered on that image:

sudo virt-install \
  --name web1 \
  --memory 2048 --vcpus 2 \
  --osinfo ubuntu24.04 \
  --disk size=20,backing_store=/var/lib/libvirt/images/noble-server-cloudimg-amd64.img \
  --cloud-init ssh-key="$HOME/.ssh/id_ed25519.pub" \
  --network network=default \
  --graphics none --import --noautoconsole

Find its address and log in:

sudo virsh domifaddr web1
ssh [email protected]

To install from an ISO instead, replace the --disk, --cloud-init and --import lines with --disk size=20 --cdrom /var/lib/libvirt/images/installer.iso and connect to the installer with Virt-Manager or a VNC viewer. Run virt-install --osinfo list to see the operating system names your version knows.

5. Manage VMs with virsh

Run these with sudo, or as a member of the libvirt group with -c qemu:///system added, so virsh talks to the system VMs rather than an empty per-user session.

virsh destroy does not delete anything; it only pulls the virtual power cord. virsh undefine without --remove-all-storage removes the VM definition but leaves its disk files.

If you prefer a graphical tool, install virt-manager on your desktop and connect to the server over SSH (qemu+ssh://user@server/system).

6. Networking: NAT or a bridge

libvirt creates a default NAT network (virbr0, 192.168.122.0/24). VMs can reach the internet, but the outside world cannot reach them unless you forward ports. That is fine for development and for VMs behind a reverse proxy on the host.

If VMs need their own addresses on your network, create a bridge. On Ubuntu, edit the netplan file (for example /etc/netplan/50-cloud-init.yaml; your file name may differ) and move the address from the physical interface to br0:

network:
  version: 2
  renderer: networkd
  ethernets:
    enp1s0:
      dhcp4: false
  bridges:
    br0:
      interfaces: [enp1s0]
      dhcp4: true

Replace enp1s0 with your interface name (ip link). Apply it with sudo netplan try, which rolls back automatically if you lose connection, then sudo netplan apply. Attach VMs with --network bridge=br0.

7. Storage pools, volumes and snapshots

libvirt stores disks in storage pools; each disk is a volume. The default pool is /var/lib/libvirt/images. To add a pool on another disk:

sudo virsh pool-define-as data dir --target /srv/vm-data
sudo virsh pool-build data
sudo virsh pool-start data
sudo virsh pool-autostart data

Create a 10 GB qcow2 volume and attach it to a VM as a second disk:

sudo virsh vol-create-as data web1-data.qcow2 10G --format qcow2
sudo virsh attach-disk web1 /srv/vm-data/web1-data.qcow2 vdb \
  --driver qemu --subdriver qcow2 --persistent

Inside the VM, format and mount /dev/vdb as normal.

Snapshots of qcow2 disks are one command:

sudo virsh snapshot-create-as web1 before-upgrade
sudo virsh snapshot-list web1
sudo virsh snapshot-revert web1 before-upgrade

A snapshot lives on the same disk as the VM, so it is not a backup. Copy disk images, or back up from inside the VM, to separate storage.

8. Performance and security basics

• Use virtio devices for disk and network; virt-install does this by default for modern guests.
• Do not overcommit memory heavily; a host that runs out of RAM kills VMs.
• Keep the host minimal: no other services on it, SSH with keys only, a firewall, and regular updates.

9. Running this on Domain India

KVM needs a server you control with hardware virtualization:

• A bare-metal server. Domain India's dedicated servers are physical machines with no virtualization layer, and the live page lists IPMI console access, which helps when you configure a bridge remotely. You run and maintain the hypervisor yourself.
• A Domain India VPS. Our VPS plans are themselves KVM virtual machines, self-managed with full root access. The live page confirms Docker, Kubernetes and LXC containers work. Running KVM inside a VPS needs nested virtualization, which we have not documented; run kvm-ok on the VPS, and ask support before you plan around it.

If you only need isolated apps rather than whole operating systems, containers on a VPS are lighter than nested VMs.

Ready to build your own virtualization host? Compare dedicated servers and VPS plans, or open a ticket to ask about nested virtualization before you buy.