Cloud Parts

Proxmox Virtual Environment

A Debian-based server platform from Vienna that runs KVM virtual machines and LXC containers, with clustering, high availability, Ceph and ZFS storage, backups and a web interface in one free ISO.

What it is
Hypervisors
Used for
Virtual machines
Type
Open source
Made by
Proxmox Server Solutions GmbH
License
AGPLv3
Facts checked
11 October 2026
Website Source code

What it is

Proxmox VE installs from an ISO onto bare metal and turns a Debian server into a virtualization host. It offers two kinds of guest: virtual machines, each a QEMU process accelerated by the kernel’s KVM and managed with qm, and Linux containers (LXC), managed with pct; since 9.1 (November 2025) a container can also be created from an OCI image. Everything is driven through one REST API whose every call is declared in JSON Schema: the ExtJS web interface is a client of it, and pvesh calls it from a root shell. There is no separate management server. Each node can manage the whole cluster, because the configuration of every guest lives in the Proxmox Cluster File System (pmxcfs), a small database-backed file system (at most 128 MiB) that Corosync replicates to all nodes in real time.

pvecm joins servers into a cluster; there’s no fixed node limit, and the docs mention production clusters of over 50 nodes. The built-in HA manager restarts VMs and containers from a failed node elsewhere after watchdog-based fencing; it needs at least three nodes for quorum and shared storage, and typical failover takes about 2 minutes. Disks can live on local ZFS, LVM or directories, or on shared NFS, CIFS, iSCSI, Fibre Channel, Ceph RBD or CephFS, and the web interface can deploy and manage a Ceph cluster on the hypervisor nodes themselves (hyper-converged, at least three nodes). Backups go through the built-in vzdump or to Proxmox Backup Server, a separate Rust product (2020) with incremental, deduplicated backups. The SDN stack adds VXLAN and EVPN zones and has been fully supported since 8.1. An import wizard, released in March 2024 and part of 8.2 that April, copies VMs from VMware ESXi 6.5 to 8.0 through ESXi’s public API and can boot them while the disks are still copying.

The current major version, 9, is based on Debian 13 “Trixie”: 9.0 shipped on August 5, 2025, and 9.2 on May 21, 2026 with Linux 7.0, QEMU 11.0, LXC 7.0, ZFS 2.4 and Ceph Squid or Tentacle. In August 2026 an arm64 build followed, officially supported on NVIDIA Grace and Vera servers. All of it is free under the AGPLv3. What Proxmox sells is a subscription per CPU socket per year (Community €120, Basic €370, Standard €550, Premium €1,100 in October 2026) that unlocks the pve-enterprise package repository and, from Basic up, support tickets. A fresh install points at that repository; without a key you switch to pve-no-subscription, which Proxmox calls suitable for testing and non-production use, and the web interface shows a “No valid subscription” warning after every login.

Why teams pick it

  • Everything in one ISO and one web interface: VMs, containers, clustering, HA, Ceph and ZFS storage, backups, firewall and SDN, with no separate management server or appliance
  • Every feature is in the free AGPLv3 code; the paid subscription buys a more tested package repository and support, priced per CPU socket rather than per core or by feature edition
  • Built from standard Linux parts (Debian, KVM, QEMU, LXC, ZFS, Ceph), so existing Linux skills and any storage Debian supports carry over
  • A practical exit from VMware: the import wizard reads VMs straight from ESXi through its API, maps most settings and can start the guest while its data is still being copied
  • Everything the GUI does is in a REST API declared in JSON Schema, scriptable from any node with pvesh

Watch out for

  • High availability needs at least three nodes and shared or replicated storage, and failover takes about 2 minutes, so Proxmox itself says HA can’t go beyond about 99.999%
  • The free path is the pve-no-subscription repository, which Proxmox says isn’t recommended for production because its packages are less heavily tested; without a key the default enterprise repository gives update errors and every login shows a subscription warning
  • Each major version follows a Debian release and is supported for about three years: Proxmox VE 8 reached end of life in August 2026, so clusters must be upgraded in place every couple of years
  • Snapshot support depends on the storage: qcow2 snapshots on file storage such as NFS block the running VM and can take minutes or hours on large disks, and snapshots on shared thick LVM (the usual iSCSI or Fibre Channel setup) arrived only as a technology preview in 9.0
  • The ESXi importer can’t read disks on vSAN or encrypted disks, and can’t carry over a VM’s virtual TPM state

Alternatives

What all hypervisors have in common →
Cloud HypervisorOpen sourceA Rust virtual machine monitor for modern cloud VMs on KVM or Microsoft’s hypervisor: virtio devices only, plus hotplug, VFIO passthrough, live migration, snapshots and Windows guests.KVM (Kernel-based Virtual Machine)Open sourceThe hypervisor built into the Linux kernel: it runs guest code directly on the CPU through `/dev/kvm`, and leaves device emulation to a user-space program such as QEMU or Firecracker.libvirtOpen sourceA stable API, daemon and `virsh` command line for managing virtual machines, storage and networks, with drivers for QEMU/KVM, Xen, LXC, Cloud Hypervisor, VMware ESX, Hyper-V and more.Microsoft Hyper-VCommercialMicrosoft’s type-1 hypervisor, built into Windows Server and Windows 10 and 11 Pro and Enterprise: a thin hypervisor under a Windows root partition, which also underpins Azure’s hypervisor and Windows features like VBS, WSL 2 and Windows Sandbox.Oracle VirtualBoxOpen sourceOracle’s free, open-source desktop hypervisor: a type-2 hypervisor that runs VMs as an application on Windows, macOS, Linux and Solaris, scriptable with `VBoxManage` and the default provider for Vagrant.QEMUOpen sourceThe open-source machine emulator and virtualizer: it emulates whole computers in software, or runs them at native speed on a hypervisor such as KVM, and supplies the virtual devices either way.VMware ESXiCommercialVMware’s bare-metal hypervisor (called ESX again since VCF 9), whose own kernel runs VMs directly on servers while vCenter adds vMotion, HA, DRS and vSAN; since Broadcom bought VMware in 2023 it is sold only as per-core subscriptions.Xen Project hypervisorOpen sourceThe open-source type-1 hypervisor from Cambridge that ran the first Amazon EC2 instances: a small microkernel-style core that boots first, with drivers and management in a privileged VM called dom0.

How it fits with other technologies

Create, start and stop VMs with `qm`, then call the REST API with `pvesh` (examples from the Proxmox VE docs; run as root on a node) · bash

# VM 300: a 4 GB IDE disk on local-lvm, an e1000 NIC and an ISO in the CD drive
qm create 300 -ide0 local-lvm:4 -net0 e1000 -cdrom local:iso/proxmox-mailgateway_2.1.iso
qm start 300

# Ask the guest to shut down, then wait at most 40 seconds for it to stop
qm shutdown 300 && qm wait 300 -timeout 40

# VM 9000 booting from an imported Ubuntu cloud image on a VirtIO SCSI disk
qm create 9000 --memory 2048 --net0 virtio,bridge=vmbr0 --scsihw virtio-scsi-pci
qm set 9000 --scsi0 local-lvm:0,import-from=/path/to/bionic-server-cloudimg-amd64.img

# The same API the web interface uses, called directly from the shell
pvesh get /nodes
pvesh set cluster/options -console html5

Where guest disks can live#

A selection of the storage types in the Proxmox VE storage docs. Shared storage is what lets running VMs live-migrate between nodes and HA restart them elsewhere.

StoragePluginLevelSharedSnapshots
ZFS (local)zfspoolfile and block (VM disks are zvols)noyes
Directorydirfilenowith qcow2 files
LVM-thinlvmthinblocknoyes
LVMlvmblockonly on a shared iSCSI, FC or SAS LUNas volume chains since 9.0
NFS and CIFSnfs, cifsfileyeswith qcow2 files
Ceph RBDrbdblockyesyes
CephFScephfsfileyesyes
Proxmox Backup Serverpbsbackups onlyyesn/a
  • BTRFS is also available as a technology preview, and ZFS over iSCSI lets a remote ZFS box serve VM disks.

See it in a real stack

No architecture uses Proxmox Virtual Environment yet. Drop it into the playground and wire it up yourself.Open playground →
Where it came from2005–2026

Martin Maurer and Dietmar Maurer founded Proxmox Server Solutions, based in Vienna, in 2005; its first product was Proxmox Mail Gateway. The virtualization project began in 2007, and Proxmox VE 0.9 came out on April 15, 2008, followed by 1.0, the first stable release, on October 29, 2008. It used KVM for virtual machines and OpenVZ for containers, with live migration for both, but clustering was limited and the interface was a simple server-generated web page. Proxmox says it was among the first companies to offer commercial support for KVM.

Version 2.0 in March 2012 rebuilt the product around three ideas that still define it: a Corosync cluster with the pmxcfs configuration file system, so every node is a master; a REST API declared in JSON Schema; and a new web interface built on Ext JS 4 that talks to that API. Ceph on the hypervisor nodes followed as a technology preview in 3.2 (2014) and ZFS support in 3.4 (2015). Version 4.0 in October 2015 replaced OpenVZ with LXC and introduced the current HA manager.

Proxmox Backup Server, written in Rust, came out in 2020. In March 2024 Proxmox added an import wizard for VMware ESXi VMs, which shipped in 8.2 that April. Version 9.0, on Debian 13, followed in August 2025, and in 2026 Proxmox VE 9.2 added dynamic load balancing of HA guests and became available for arm64 servers.

Why it’s built this way

A Debian distribution built from stock open-source parts
Proxmox uses KVM, QEMU, LXC, ZFS and Ceph instead of writing its own hypervisor or storage, so any storage technology Debian supports can hold VM images and users aren’t tied to one vendor’s stack.
Multi-master clustering on pmxcfs instead of a management server
Corosync replicates the configuration file system to every node, so any node can manage the whole cluster and there is no central server to fail or to license separately.
API first, with the GUI as a client
Declaring the whole REST API in JSON Schema in 2.0 let hosting companies and tools integrate Proxmox VE, and made it possible to replace the server-generated pages with a browser application.
Sell repository access and support, not feature editions
All features stay in the AGPLv3 code. Subscriptions are priced per CPU socket and buy the more tested enterprise repository and support, so a home lab and a paying customer run the same software.

Timeline

  1. 2005

    Martin and Dietmar Maurer found Proxmox Server Solutions

  2. 2008

    0.9 in April, the first public release; 1.0 in October with KVM and OpenVZ

  3. 2012

    2.0 in March: Corosync cluster, pmxcfs, REST API and a new web interface

  4. 2015

    4.0 in October replaces OpenVZ with LXC and adds the HA manager

  5. 2020

    First stable Proxmox Backup Server, written in Rust

  6. 2024

    ESXi import wizard in March, shipped in 8.2 in April

  7. 2026

    9.2 in May on Linux 7.0; first arm64 release in August

Sources17 · facts checked 11 October 2026

Improve this page on GitHub