Proxmox VE: the free hypervisor that runs my whole lab
Every service in this lab runs as a virtual machine. The firewall, the DNS resolver, the photo server, the AI inference stack, the monitoring platform, the password manager — all of them are isolated guest systems living on top of two physical servers. That’s only possible because of a single piece of software sitting underneath everything: Proxmox VE.
If you’re new to homelabbing and trying to figure out how serious setups actually work, Proxmox is one of the most important pieces of the puzzle to understand. This post is the explanation I wish I’d had when I started.
What Proxmox VE actually is
Proxmox VE (Virtual Environment) is a Type-1 hypervisor — meaning it runs directly on bare metal, not on top of another operating system. You install it on a server, it takes over, and then everything else you run lives inside virtual machines or containers that it manages.
The core technology underneath is KVM (Kernel-based Virtual Machine), which has been part of the Linux kernel for well over a decade and is what the major cloud providers build on. Proxmox layers a polished web interface, a REST API, clustering support, and a backup system on top of that foundation.
It’s free and open source. The community edition has no meaningful functional restrictions. There’s an optional paid support subscription that funds development, but the software itself is fully capable without it.
The problem it solves
Before hypervisors, if you wanted to run five different services, you needed five different physical machines (or you crammed them all onto one and hoped they didn’t interfere with each other). Virtualization changes that completely.
With Proxmox, one physical server can run dozens of isolated guest systems simultaneously. Each guest thinks it’s on its own hardware. They don’t share processes, they can’t see each other’s memory, and if one crashes or gets misconfigured, it doesn’t take the others down with it. You can snapshot a VM before making changes, roll it back if something breaks, and migrate it to a different host with a few clicks — all while it’s running.
For a homelab, this is transformative. A single powerful server can run your network firewall, your media stack, your monitoring stack, and your development environments, all isolated from each other. You don’t need a different physical machine for each role.
The commercial equivalent
In enterprise IT, the dominant player in this space has historically been VMware (now Broadcom). VMware vSphere with ESXi is the gold standard in data centers — it’s what most sysadmins and infrastructure engineers will encounter in a job environment. It’s mature, well-documented, and deeply integrated into the enterprise toolchain.
The catch: after Broadcom acquired VMware in late 2023, the licensing model changed dramatically. What was once accessible — even for labs and small organizations — became expensive in ways that made personal or small-scale use untenable. Per-core licensing, bundled suite requirements, and the elimination of standalone perpetual licenses pushed many people to look elsewhere.
⚠️ Unverified: The specific details of Broadcom’s post-acquisition VMware licensing changes — exact pricing tiers, elimination of specific SKUs, and timing — are general industry knowledge. Verify current terms at broadcom.com before making purchasing decisions.
Nutanix is another enterprise option — it combines hypervisor, storage, and management into a converged platform, and it’s excellent in enterprise environments, but it’s even further outside reach for personal use.
The Broadcom situation is actually what pushed a lot of homelab operators toward Proxmox. For running a personal lab, the tradeoffs became obvious fast.
The self-hosted alternatives worth knowing
When you start looking at open hypervisors, a few names come up:
Proxmox VE — the one this post is about. Based on Debian Linux, uses KVM for full virtualization and LXC for lightweight containers. Has a genuinely good web interface, a mature API, a Terraform provider, clustering via Corosync, and an active community. The learning curve is real but the documentation is solid.
XCP-ng — a community fork of Citrix XenServer, which itself is enterprise-grade virtualization software. XCP-ng is polished, pairs well with the Xen Orchestra management interface, and is a legitimate alternative to Proxmox. The Xen hypervisor is different from KVM under the hood, but the operational concepts are similar. It’s particularly popular in the enterprise homelab crowd.
Microsoft Hyper-V — Windows Server includes Hyper-V, and there’s also a free standalone Hyper-V Server edition.
⚠️ Unverified: The current availability and support status of the free standalone Hyper-V Server product may have changed — verify with Microsoft directly.
For homelabbers who live in the Windows ecosystem, Hyper-V works, but the management story is more Windows-centric and the open-source community ecosystem around it is smaller.
Plain KVM / libvirt — you can absolutely run virtual machines directly on any Linux system without a dedicated hypervisor layer. Tools like virt-manager and virsh give you a management interface. This is the most lightweight option but you’re giving up the web interface, the cluster management, the built-in backup tooling, and the Terraform integrations. It’s a fine way to learn KVM fundamentals, but it doesn’t scale the way Proxmox does.
The reason Proxmox wins for most serious homelabbers is the combination: it’s free, it runs KVM (which is directly transferable knowledge to cloud environments), it has a real cluster model, and the API is first-class. Everything in this lab is provisioned via Terraform using the bpg/proxmox provider, which means the entire VM inventory is code — repeatable, reviewable, version-controlled.
How it fits a serious homelab
This lab runs two physical servers, both Dell rack servers, forming a Proxmox cluster. They share the same management plane — one web interface, one API endpoint, a single view of all the VMs across both hosts.
Across those two hosts, the cluster currently runs about 38 virtual machines and has the capacity for significantly more. They range from tiny two-core, two-gigabyte services to large memory-intensive workloads with tens of virtual cores and hundreds of gigabytes of RAM. The hypervisor handles the scheduling.
A few things Proxmox makes routine that would otherwise be painful:
Templates and cloud-init. Every new Ubuntu VM comes from a cloud-init template — a golden image that’s been configured once and converted to a template. Provisioning a new VM takes seconds and produces a predictable, consistent base system. Cloud-init injects the hostname, SSH keys, and network configuration on first boot. Terraform handles the provisioning call. The whole flow from code to running VM is a few minutes.
Live migration. VMs can be moved between physical hosts while running. When one host needs maintenance, or the load is unbalanced, or you want to test something — you migrate, you don’t rebuild. This was enterprise-only behavior a decade ago.
Snapshots. Before any significant change — a Terraform module update, a tricky configuration, a kernel upgrade — a snapshot of the VM takes seconds. If the change breaks something, you roll back and you’re immediately back to a known-good state. This removes a huge amount of hesitation from experimentation.
Proxmox Backup Server. PBS is a companion product (also free) that handles VM and container backups with deduplication. This lab runs PBS as a VM on the cluster itself, backed by an NFS datastore on the storage server. Incremental backups, retention policies, and restore tests are all built in.
The API. Everything Proxmox can do through the web interface, it can also do through a REST API. The Terraform provider wraps that API, which means VM lifecycle management — create, configure, start, stop, destroy — is all handled in infrastructure code. The entire VM inventory is in a Git repository. When you need to rebuild something, you run terraform apply, not click through a wizard.
Clustering with Corosync. The two hosts form a cluster with Corosync handling the membership and fencing. This is the same clustering technology used in enterprise Linux HA configurations. Building familiarity with it in a homelab directly maps to how it shows up in production environments.
ZFS underneath
Proxmox has first-class ZFS integration, and running ZFS for VM storage is one of the better decisions you can make in a homelab.
ZFS gives you copy-on-write semantics (which makes snapshots nearly instant and space-efficient), checksumming on every block (which means silent data corruption gets detected and reported rather than silently corrupting your VMs), and a flexible pool model for managing storage across drives.
Each host in this lab uses ZFS mirror pools for the primary VM storage — two drives mirrored, so a single drive failure doesn’t lose anything. The OS for the hypervisor itself lives on a separate mirrored NVMe pair. Understanding ZFS pool design — mirror vs RAIDZ, vdev layout, when to use what — is worth a full post on its own, and it’s a skill that transfers to any serious storage context.
Who should bother with this
Proxmox is worth learning if:
- You want to understand virtualization the way it actually works in infrastructure jobs. KVM, QEMU, cloud-init, Corosync — these are real-world technologies.
- You want to run multiple services without managing multiple physical machines.
- You like the idea of everything being infrastructure-as-code. The Terraform provider is mature and the API is comprehensive.
- You’re self-hosting enough services that isolation, snapshots, and easy rebuilds genuinely matter.
It’s probably overkill if you’re running one or two services on a Raspberry Pi or a small NUC. For that scale, Docker on a single machine or a NAS with Docker support is simpler and faster to get running. Proxmox is the right call when you’re thinking at the scale of a rack, or when your goal is specifically to learn the technology.
The learning curve is real. You will spend time understanding how bridged networking works, how cloud-init templates are built, how storage pools map to actual drives, and what Corosync is actually doing. None of that is wasted time — it’s all knowledge that pays back in a professional environment.
The short version
Proxmox VE is a free, enterprise-quality hypervisor that runs on your hardware and manages virtual machines through a polished web interface and a full-featured API. It’s built on KVM, clusters natively, integrates with Terraform, and supports ZFS. The commercial equivalent is VMware, which became significantly harder to justify for personal or small-scale use after the Broadcom acquisition.
For anyone building a serious homelab — one that’s meant to teach real skills, not just run services — Proxmox is where the foundation goes. Everything else runs on top of it.
Comments
No comments yet — be the first.