Guide
Migrating from VMware to Proxmox: what to settle before the first move
Vendor statements: Proxmox VE 8.2 (import wizard), ESXi 6.5 to 8.0 (tested by the vendor). No versions of our own.
Before the first virtual machine moves from VMware to Proxmox VE, settle seven things: a complete inventory, the dependencies between systems, the storage design, the network, the guest systems, the licence and contract timeline, and a pilot with a defined way back. For each point, this article shows what goes on the list and which limits the vendors document.
It describes the preparation, not the move itself. Every vendor statement is listed with its source and the date we checked it at the end.
1. Why the preparation sets the timeline
Since version 8.2 (released on 24 Apr 2024), Proxmox VE has shipped an import wizard for virtual machines from VMware ESXi. According to the Proxmox documentation, the import was tested with ESXi 6.5 to 8.0 (documentation as of 27 May 2026). Copying a VM is therefore rarely the hard part.
It gets hard at the points the import does not see: a system three others depend on; a licence bound to a MAC address; a backup only VMware knows about; a maintenance window nobody agreed. You find these points either beforehand, calmly, or afterwards, under time pressure.
2. Inventory: what goes on the list
One row per virtual machine. You will need these columns later for the order, the target design and the maintenance windows:
| Column | What for |
|---|---|
| Name, purpose, person responsible on the business side | Who signs off the test after the move |
| Operating system and version | Drivers, vendor support, licence binding |
| vCPU, memory, disk capacity (used and provisioned) | Capacity of the target cluster |
| Firmware (BIOS or UEFI), vTPM yes/no | Settings of the target VM; vTPM see section 5 |
| Datastore type (local, SAN/NAS, vSAN), encryption yes/no | Whether the direct import works at all (section 3) |
| Snapshots yes/no | Snapshots slow the import down considerably (section 3) |
| Networks and IP configuration (static or DHCP reservation) | Section 4 |
| VMware features in use (HA, vMotion, snapshots as backup, replication) | What needs a counterpart on Proxmox VE |
| Backup: tool, frequency, last successful restore | Section 7 |
| Permitted downtime and possible maintenance windows | Order and waves |
For the technical columns, an export from vCenter is enough. An example with VMware PowerCLI (generic file names):
# One row per VM: name, state, CPU, RAM, disk space, guest OS, firmware
Get-VM | Select-Object Name, PowerState, NumCpu, MemoryGB, ProvisionedSpaceGB,
@{N='GuestOS'; E={$_.Guest.OSFullName}},
@{N='Firmware'; E={$_.ExtensionData.Config.Firmware}} |
Export-Csv -Path vm-inventory.csv -NoTypeInformation -Encoding UTF8
# Every snapshot with age and size
Get-VM | Get-Snapshot | Select-Object VM, Name, Created, SizeGB |
Export-Csv -Path vm-snapshots.csv -NoTypeInformation -Encoding UTF8
The business columns (purpose, responsible person, permitted downtime) are in no export. You fill them in with the business units.
3. Dependencies and storage
Dependencies. For every system, note what it depends on and what depends on it: database, directory service, DNS, file shares, licence server, interfaces to external partners. Systems that work closely together move in the same wave. A system many others depend on does not move first.
Storage: what the import cannot do. The Proxmox documentation names these limits for the import from ESXi:
- VMs with disks on vSAN cannot be imported directly.
- Encrypted VM disks (for example via a storage policy) cannot be imported.
- With snapshots, the import can be considerably slower.
Every affected VM needs a workaround in the plan: consolidate snapshots beforehand, move disks to another datastore beforehand, or remove the encryption beforehand, each within the maintenance window.
Storage: the target. For shared storage in the cluster, Proxmox recommends Ceph. Among other things, the documentation names:
- at least three servers, as similar as possible, and at least three monitors for quorum;
- at least 10 Gbit/s of network bandwidth used exclusively for Ceph traffic;
- a host bus adapter instead of a RAID controller, SSDs preferred, disks of the same size and evenly distributed per server;
- at least 8 GiB of memory per OSD (per Ceph disk), plus headroom for recovery and rebalancing.
Whether your existing servers meet these points belongs in the hardware inventory. If they do not, the decision “new hardware or a different storage design” is part of the target design, not a surprise in the project.
4. Network
- New MAC addresses. The VM’s network card changes with the move. Adjust DHCP reservations beforehand. On Windows, a static IP configuration can stay attached to the old card; the Proxmox documentation recommends noting it down beforehand.
- Bridges instead of port groups. Proxmox VE connects VMs to the network via Linux bridges; VLANs can be set in several places. Translate every port group into a bridge or a VLAN tag and record the mapping in a table.
- Separate networks for the cluster. The Ceph documentation advises physically separate networks for Ceph cluster traffic, for public Ceph and VM traffic, and for cluster communication (Corosync). That is a question for your switches and cabling, to settle before you order.
- Licences tied to the network. Some software binds its licence to a MAC address or to hardware characteristics of the VM. Ask the vendor before the address changes.
5. Preparing the guest systems
From the Proxmox documentation, for every VM before the import:
- Remove VMware Tools. The old hypervisor’s guest tools should be uninstalled before the move.
- Note the network configuration (see section 4).
- vTPM and disk encryption. If the keys of a disk encryption live in the virtual TPM, the documentation advises disabling the encryption beforehand.
- Shut down the source VM. Even with the live import, which shortens the downtime, the VM stays switched off on ESXi; so there is always some downtime.
After the import:
- Disk bus. The documentation recommends SCSI with the “VirtIO SCSI single” controller. Windows VMs need extra steps for this so that the boot disk starts with the VirtIO drivers. Plan these steps into the pilot, not for the tenth VM.
- CPU type. It should match the underlying hardware as closely as possible; with mixed hardware in the cluster, the documentation names the generic “x86-64-v…” models.
6. Licences and contract timeline
The VMware side. In December 2023, VMware stopped selling perpetual licences and renewing the associated support and subscription contracts (SnS); since then the products have been available as subscriptions.
For existing perpetual licences, Broadcom describes in its knowledge base what applies after the SnS contract ends: ESXi and vCenter keep running, running VMs stay on, and the features of the licence edition remain unlocked. But there is no longer access to new patches and security updates, and no support requests.
For your timeline this means: the date you plan for is the end of your current subscription or SnS contract. Count back from there: inventory, target design, hardware lead time, pilot, waves. What your own contract allows after that date is for your legal adviser or licensing consultant to settle; we do not review contracts.
The Proxmox side. Proxmox VE is open source (AGPLv3). The subscription is charged per occupied physical CPU socket per year. It gives access to the enterprise repository, which the Proxmox documentation recommends for production use. According to the documentation, the repository without a subscription is meant for testing and non-production use. For your budget, what counts is the number of CPU sockets in the target cluster, not the number of VMs. Prices are on the vendor’s website.
Guest systems. The licences of your operating systems and applications do not change automatically when the hypervisor changes. Check licences bound to hardware, a host or a MAC address (section 4).
7. Backup and monitoring
Proxmox VE backs up VMs with the built-in tool vzdump or with Proxmox Backup Server (with deduplication). Not every backup software you use under VMware supports Proxmox VE. Ask your vendor and record the answer with a date.
Before the pilot, decide how the first VM on Proxmox VE will be backed up, and restore it once from that backup. A backup you have never restored from is a promise, not proof.
The same goes for monitoring: which alerts come from vCenter today, and who gets them after the move?
8. Pilot and way back
Pilot. For the first move, choose two or three systems that are not critical but still typical: at least one Windows and one Linux VM, ideally one with snapshots or a static IP. For each, record:
- how long the import took and how long the VM was unreachable;
- which steps were needed after the import (drivers, network, CPU type);
- whether the responsible person on the business side signed off the functional test;
- whether backup and restore worked on the new platform.
Plan the waves from these measurements, not from estimates.
Way back. Define it before the first move:
- The source VM stays on ESXi, switched off, until sign-off. It is only deleted afterwards.
- Write down the point in time up to which you will go back. After the move, the system writes data on the new platform. Anyone going back later has to take that data along or lose it.
- If you change DNS records, lower their TTL beforehand so that a way back does not hang on cached answers.
- A way back nobody has rehearsed is a plan. Rehearse it once in the pilot.
9. Checklist
Inventory:
- One row per VM with every column from section 2, business columns filled in with the business units.
- Dependencies noted per system; waves formed from them.
- VMs on vSAN, with encrypted disks or with snapshots marked, each with a workaround.
Target design:
- Storage design decided; with Ceph: three or more servers, a dedicated network with at least 10 Gbit/s, HBA instead of a RAID controller.
- Network: mapping port group → bridge/VLAN tag; separate networks for Ceph and Corosync planned.
- Backup and monitoring for Proxmox VE decided.
Dates and licences:
- End of the VMware subscription or SnS contract known; timeline counted back from there.
- Number of CPU sockets in the target cluster determined for the Proxmox subscription.
- Licences bound to hardware or a MAC address settled with the vendor.
Per VM before the move:
- VMware Tools removed, network configuration noted, DHCP reservation adjusted.
- vTPM-bound encryption disabled where present.
- Maintenance window agreed; source VM kept until sign-off.
Pilot:
- Two or three typical, non-critical systems moved; duration, follow-up work and sign-off recorded.
- Backup and restore carried out once on Proxmox VE.
- Way back rehearsed once.
Where this leads
We run Proxmox VE and Ceph ourselves in production. We run our own hardware in a data centre in Munich. Fortlauf is our product. We built it, and we run it on our own hardware. What we measured there is listed with dates on the page Where your data lives.
How we structure a move from the inventory to the handover is described under VMware → Proxmox migration.
If Ceph provides the storage after the move: Encrypting Ceph OSDs after the fact and Testing database restores.
German version: VMware zu Proxmox migrieren: was Sie vor dem ersten Umzug klären.
Sources
All read on 12 Oct 2026.
- Proxmox VE Wiki: Migrate to Proxmox VE, last changed on 27 May 2026. Import wizard, tested ESXi versions, limits (vSAN, encrypted disks, snapshots), guest preparation, network, disk bus, CPU type, backup.
- Proxmox: Proxmox Virtual Environment 8.2, press release for the version of 24 Apr 2024. Import wizard for VMware ESXi.
- Proxmox VE documentation: Deploy Hyper-Converged Ceph Cluster. Recommendations for servers, network, disks and memory.
- Proxmox VE Wiki: Package Repositories, last changed on 21 May 2026. Enterprise repository and repository without a subscription.
- Proxmox: Proxmox VE Subscriptions. Counted per CPU socket, AGPLv3 licence.
- Broadcom Knowledge Base, article 429208: Perpetual License Functionality After Support and Subscription (SnS) Expiration.
- Redmond Channel Partner: Broadcom Kills VMware Perpetual Licenses, 13 Dec 2023.
The matching serviceVMware to Proxmox migration