Release notes

Virtainer Free

Stable releases, newest first. Release candidates are folded into the stable release they led to. For how to install or update, see theinstall guideand OS updates.

1.1.0 · October 6, 2026

The machines themselves: a guest agent, shell, snapshots and first-boot setup that Virtainer now provides end to end, validated through forty release candidates on real hardware.

Introducing Virtainer Free 1.1

Inside the VM

  • A Virtainer-built guest agent, the Virtainer Guest Agent (VGA), is installed into every classic VM at first boot from the host's own copy. Nothing is downloaded and it does not depend on the distribution's packages.
  • The Shell tab opens a terminal over vsock, so it works without guest networking or SSH, supports several sessions, and is recorded in the audit log. The serial console becomes the Rescue console for machines that are not healthy.
  • The instance list flags a VGA that is out of date, with an upgrade button next to the version. Select several instances to update them together, or use Update VGAs for all.
  • Hot-added CPUs and memory are brought online by the agent on every supported distribution, including images that leave them offline.
  • A VM that has not reported in yet shows No boot signal instead of Failed.

Images and first boot

  • Every image you add is read before it boots, so first-boot configuration is written for what the image really contains: Alpine works, users get a login shell the image has, and serial auto-login follows the image's init and getty.
  • An image that can only boot through BIOS is refused when you add it.
  • Local images keep their distribution after a reload, and adding an image by URL shows progress under the distribution you chose.

Creating VMs

  • The create page and the VM template form are one set of settings. Required fields come first, followed by settings that show their current value; templates can store everything except identity.
  • A template is pinned to one image, can carry several network interfaces, and can use a saved SSH public key.
  • Saved SSH public keys are managed on the same page as API keys, and can be saved from the create page without losing the form.
  • Disk space is checked before creation, and memory is entered in 256 MB steps.

Snapshots

  • Snapshots are qcow2 internal snapshots: taken in seconds for running and stopped VMs alike, organised as a tree, and rolling back restores the whole machine, including its disks and network interfaces.
  • Each snapshot states its consistency honestly, and a filesystem-consistent snapshot explains why when the guest agent is not ready.
  • Snapshots travel with a VM when you export it.

Networking

  • A network interface is a resource of its own: create it, attach it to a VM (hot-plug works on a running VM), detach it back to the pool, or delete it.
  • The real state of host networking and the stored configuration no longer drift apart silently, and a network change that is not confirmed is rolled back.
  • The management address can change under you: the page follows to the new address.

Host and updates

  • The host image is rebuilt from the latest Fedora every time, and carries the firmware that servers and small-form-factor machines commonly need. Bluetooth drivers are no longer loaded.
  • An upgrade can no longer fail silently because of a doubled mount on /boot.
  • New hosts refuse root login over SSH, the SSH login policy is independent of the password setting, and a host no longer suspends when the lid is closed.
  • A host has a single administrator account.
  • The sidebar is split into VM and AppVM modules, with AppVM marked Preview.

Upgrading

  • Existing hosts update from OS Updates in the console.
  • A host still on 1.0.0 needs systemctl mask boot.automount run once at its console before updating.

1.0.0 · September 18, 2026

The first 1.0 release: a complete single-host appliance for VMs and AppVMs, with live snapshots, a built-in image catalogue and an image-based update path.

Highlights

  • Snapshots of running VMs, taken as a single transaction across all disks, with a three-level consistency contract that never claims more than it can prove.
  • AppVMs gained a runtime filesystem and concurrent exec, and container runtimes now work inside them.
  • The image catalogue takes a registry shape with its own management page, and a template can fetch a new upstream build when one is available.
  • A VM whose hypervisor process is merely not answering is no longer destroyed.
  • The first-run wizard works on machines with no spare disk, and the SSH key it installs now works.
  • The host is built as a single image, and upgrades come from Stable or RC channels.