Migration guide · 2026
VMware to Proxmox migration: a practical 2026 guide for SMB IT admins
Broadcom's acquisition of VMware replaced perpetual licenses with per-core subscriptions and a 16-core-per-CPU minimum. For many small and mid-market teams that meant renewal quotes 2–5× higher than before. This guide walks through planning and executing a move to Proxmox VE without the consultant invoice.
Why teams are leaving VMware in 2026
The licensing model changed, not just the price. VMware vSphere is now sold through bundles like VMware vSphere Foundation (VVF) and VMware Cloud Foundation (VCF), priced per physical core with a minimum of 16 cores billed per CPU socket. A two-socket host with 10-core CPUs is still billed as 32 cores. Multiply that across a cluster and a previously predictable maintenance renewal becomes a five- or six-figure annual subscription.
Proxmox VE is the most common like-for-like replacement for on-prem VMware. It is an open-source virtualization platform built on KVM and LXC, with a built-in web UI, clustering, high availability, and software-defined storage via Ceph or ZFS. There is no per-core license; an optional enterprise support subscription runs roughly €340 per CPU socket per year. The hardware you already own is reused.
Step 1 — Inventory your estate
You cannot plan a migration you cannot see. Export your full inventory before anything else. The standard tool is RVTools, a free utility that connects to vCenter and dumps every VM's configuration. Export the vInfotab to CSV — it contains VM name, power state, vCPU count, memory, provisioned storage, guest OS, host and cluster. If you don't use RVTools, a vCenter VM list exported to CSV works too.
With the inventory in hand, the first wins are free: identify powered-off VMs and orphaned templates. These are decommission candidates that reduce both your migration scope and your licensed core count.
Step 2 — Map every VM to a target
Not every VM should move the same way. A useful default mapping:
- Standard production VMs → Proxmox VE, like-for-like KVM.
- Large or database VMs → Proxmox with shared storage (Ceph or ZFS), with IOPS validated before cutover.
- Small stateless Linux services → low-risk pilot candidates; also good future containerization targets.
- Legacy/end-of-life OSes → migrate as-is to keep the lights on, then schedule an upgrade or re-platform.
- Powered-off VMs → decommission rather than migrate.
Step 3 — Plan in phases
Big-bang cutovers fail. A phased rollout de-risks the move:
- Phase 0 — Discovery & decommission: document networking and dependencies, retire dead VMs.
- Phase 1 — Pilot: migrate a handful of small, stateless Linux VMs to prove the Proxmox cluster, tooling and your cutover runbook end to end.
- Phase 2 — Bulk production: move standard VMs in batches by cluster and maintenance window.
- Phase 3 — Complex & stateful: databases, large VMs, clustered apps and legacy OSes last, with tested backups and a rollback plan.
Step 4 — Handle the gotchas
The migrations that go wrong usually trip on the same handful of issues:
- VirtIO drivers for Windows: Windows guests need VirtIO storage and network drivers injected before or during cutover, or they boot to a 0x7B blue screen on KVM.
- VMware Tools → qemu-guest-agent: remove VMware Tools and install qemu-guest-agent for graceful shutdown, IP reporting and quiesced backups.
- Storage: VMFS datastores don't move. Plan Ceph (hyperconverged) or ZFS/NFS and validate IOPS for database workloads.
- Networking: recreate vSphere Distributed Switch port groups and VLAN tags on Proxmox Linux bridges or the SDN stack. Document VLANs before you start.
- HA/DRS: Proxmox HA with a 3+ node corosync quorum replaces vSphere HA. There is no live DRS auto-balancing — plan manual or scripted rebalancing.
- Backups: Veeam now supports Proxmox VE, and Proxmox Backup Server is the native option. Re-establish 3-2-1 backups and test a restore before decommissioning ESXi.
Step 5 — Build the cost-savings case
To get budget approval you need a number. The simple model: estimate your billable cores (sockets × max(actual cores, 16)), multiply by the VMware per-core list price (VCF lists around $350/core/year), and compare to Proxmox's optional support subscription per socket. For most SMB clusters the result is a 60–90% reduction in annual virtualization licensing, on hardware you already own. Always replace the list price with your actual renewal quote for the real figure.
Do it in minutes instead of weeks
Everything above — the inventory parsing, target mapping, phasing, savings math and risk checklist — is exactly what VMExit automates. Upload your RVTools export and get a tailored exit plan instantly.