virt-install Command in Linux: Create KVM VMs from the CLI

Confirm virt-install is installed and see which --osinfo values libosinfo knows on your host.

Run the command:

virt-install --version

Sample output:

5.1.0

List short OS IDs you can pass to --osinfo:

virt-install --osinfo list | head -8

Sample output:

almalinux10
almalinux9
almalinux8
almalinux-kitten10
alpinelinux3.23
alpinelinux3.22
alpinelinux3.21
alpinelinux3.20

Use osinfo-query os for the full table with distro names and versions.

virt-install references network names defined in libvirt — usually default on a fresh KVM host.

Run the command:

virsh net-list --all

Sample output:

Name      State    Autostart   Persistent
--------------------------------------------
 default   active   yes         yes

Pass network=default to attach the NAT network, or bridge=br0 when the guest should share a host bridge.

If default is missing or inactive on RHEL-family hosts:

sudo virsh net-autostart default
sudo virsh net-start default

When the network does not exist, verify or reinstall libvirt-daemon-config-network.

--dry-run --print-xml validates options and prints libvirt XML without defining the guest. Include the same install method (--cdrom, --location, --pxe, --import, or --install) you plan to use — --dry-run skips changes but still runs normal validation. With --location, virt-install may fetch installation metadata or boot files for OS detection.

Run the command:

virt-install --connect qemu:///system \
  --dry-run --print-xml \
  --name demo-test \
  --memory 2048 --vcpus 2 \
  --osinfo rocky10 \
  --location https://download.rockylinux.org/pub/rocky/10/BaseOS/x86_64/os/ \
  --disk size=5,format=qcow2 \
  --network network=default \
  --graphics vnc \
  --noautoconsole

Sample output:

<domain type="kvm">
  <name>demo-test</name>
  <memory>2097152</memory>
  <vcpu>2</vcpu>
  <os>
    <type arch="x86_64" machine="q35">hvm</type>
  </os>
  <devices>
    <emulator>/usr/libexec/qemu-kvm</emulator>
    <disk type="file" device="disk">
      <driver name="qemu" type="qcow2" discard="unmap"/>
      <source file="/var/lib/libvirt/images/demo-test.qcow2"/>
      <target dev="vda" bus="virtio"/>
    </disk>

On a KVM-capable host the root element is normally <domain type="kvm">. If you see type="qemu", the host is falling back to software emulation — enable hardware virtualization in firmware and confirm the kvm module is loaded.

If the default storage pool is too small, shrink --disk size= or free space under /var/lib/libvirt/images/.

Match --osinfo to the OS on the ISO — Rocky media needs rocky10, not an RHEL profile. Linux paths are case-sensitive; use the exact filename on your host.

Run the command:

sudo virt-install --connect qemu:///system \
  --name rocky10-guest \
  --memory 4096 --vcpus 2 \
  --osinfo rocky10 \
  --cdrom /var/isos/Rocky-10.2-x86_64-dvd.iso \
  --disk size=20,format=qcow2 \
  --network network=default \
  --graphics vnc \
  --console pty,target_type=serial

VNC commonly listens on 127.0.0.1 only. Find the display with virsh vncdisplay rocky10-guest, then connect through local virt-viewer, an SSH tunnel, or a libvirt connection over SSH — you do not need to expose VNC publicly. --cdrom boots the ISO; use --location when virt-install should fetch kernel/initrd and start the installer automatically.

--location points at a distro install tree. virt-install pulls vmlinuz and initrd, then hands off to the network installer.

Run the command:

sudo virt-install --connect qemu:///system \
  --name rocky10-netinst \
  --memory 2048 --vcpus 2 \
  --osinfo rocky10 \
  --location https://download.rockylinux.org/pub/rocky/10/BaseOS/x86_64/os/ \
  --disk size=10 \
  --network network=default \
  --graphics vnc --noautoconsole

Use a mirror URL that matches your architecture. For kickstart, point inst.ks= at an HTTP, HTTPS, NFS, or other installer-accessible URL:

--extra-args "inst.ks=https://server.example.com/kickstart/rocky10.cfg"

Or inject a host-local file into the installer initrd:

--initrd-inject /path/to/rocky10.cfg \
--extra-args "inst.ks=file:/rocky10.cfg"

Confirm the profile exists locally with virt-install --osinfo list before using --install rocky10 — an older osinfo database can lack newer distributions even when virt-install supports --install.

RHEL-family hosts recommend file-backed disks as volumes in a directory storage pool. /var/lib/libvirt/images is labeled for libvirt by default; custom paths need equivalent SELinux/sVirt handling — changing ownership to qemu alone may not fix access.

Define a pool, then reference it from virt-install:

sudo mkdir -p /disks
sudo virsh pool-define-as vm-disks dir --target /disks
sudo virsh pool-start vm-disks
sudo virsh pool-autostart vm-disks

Run the command:

sudo virt-install --connect qemu:///system \
  --name app-vm \
  --memory 8192 --vcpus 4 \
  --osinfo rocky10 \
  --cdrom /var/isos/Rocky-10.2-x86_64-dvd.iso \
  --disk pool=vm-disks,size=40,format=qcow2 \
  --network bridge=br0 \
  --graphics vnc --autostart

--autostart marks the persistent domain to start automatically during host boot. When you must use a raw path=, verify parent-directory traversal, libvirt/QEMU permissions, and SELinux contexts — not just Unix ownership.

Repeat --network for each NIC. Mix NAT and bridged attachments when the guest needs management and data-plane separation.

Run the command:

sudo virt-install --connect qemu:///system \
  --name dual-nic-vm \
  --memory 4096 --vcpus 2 \
  --osinfo rocky10 \
  --location https://download.rockylinux.org/pub/rocky/10/BaseOS/x86_64/os/ \
  --disk size=20 \
  --network network=default \
  --network bridge=br0 \
  --graphics vnc --console pty,target_type=serial

List host bridges with ip link show type bridge before choosing bridge=.

When you launch from SSH, --noautoconsole returns control immediately. Finish the install through VNC or virsh console.

Run the command:

sudo virt-install --connect qemu:///system \
  --name headless-install \
  --memory 2048 --vcpus 2 \
  --osinfo rocky10 \
  --location https://download.rockylinux.org/pub/rocky/10/BaseOS/x86_64/os/ \
  --disk size=10 \
  --network network=default \
  --graphics none \
  --console pty,target_type=serial \
  --extra-args "console=ttyS0" \
  --noautoconsole

Sample output:

Domain installation still in progress. You can reconnect to
the console to complete the installation process.

Reconnect with virsh console headless-install. Press Ctrl+] to leave the serial session.

--noautoconsole detaches virt-install from the console and normally lets the command exit quickly. For a multi-stage --location or --cdrom install, the guest may be left shut off when installation completes. Start it with virsh start headless-install, or add --wait -1 when you want virt-install to remain running through the post-install reboot.

Skip the installer when you already have a configured bootable disk image or clone — not a generic cloud image that still needs cloud-init credentials.

Run the command:

sudo virt-install --connect qemu:///system \
  --name imported-vm \
  --memory 2048 --vcpus 2 \
  --import \
  --disk path=/var/lib/libvirt/images/existing.qcow2,format=qcow2 \
  --osinfo rocky10 \
  --network network=default \
  --graphics vnc --noautoconsole

The disk file must exist before you run the command. The new domain name must be unique — no virsh undefine is required when importing under a different name. When reusing an existing domain name, shut down and undefine that domain deliberately, and confirm whether its storage or UEFI NVRAM must be retained.

For cloud images that ship without a local login, add --cloud-init (optionally with root-ssh-key=, clouduser-ssh-key=, or user-data=). Plain --cloud-init creates temporary NoCloud metadata, generates a root password, prints it, and disables cloud-init after the first boot.

--pxe configures network boot for the first boot only. The selected network must reach DHCP, TFTP/iPXE, and boot configuration — use a host bridge when your PXE server lives on the physical LAN. inst.ks= belongs in the PXE boot entry the server serves, not in --extra-args (that flag is for --location installs).

Run the command:

sudo virt-install --connect qemu:///system \
  --name pxe-guest \
  --memory 4096 --vcpus 2 \
  --osinfo rocky10 \
  --pxe \
  --disk size=20 \
  --network bridge=br0 \
  --graphics vnc \
  --console pty,target_type=serial

Use network=default only when that libvirt network is configured to provide or forward PXE services. See KVM PXE netboot for the full server setup.