Windows Server 2025 → considering a migration to Proxmox VE

Pascal.ch

New Member
Jan 27, 2026
3
0
1
Hello,

I’m looking for feedback from the community regarding an architecture decision I’m currently considering after encountering several issues with Windows Server 2025.

Context​

The server was initially installed with Windows Server 2025 bare-metal, hosting the following roles:
  • AD DS
  • DNS
  • DHCP
  • Application services
After the promotion to domain controller, I started facing several blocking issues, including:
  • Essential Windows services not starting after reboot (e.g. Defender)
  • Malfunctions in services relying on COM / DCOM (e.g. Active Image Protector)
  • Security hardening automatically applied by AD DS affecting applications not designed to run on a DC
  • System restore difficulties (drivers / storage) despite valid backups
Despite rebuilds and clean restores, the environment proved insufficiently reliable for production.

Current architecture​

HPE ProLiant DL20 Gen11
└── Windows Server 2025 (bare-metal)
├── AD / DNS / DHCP
└── Applications

Target architecture​

I’m considering moving to Proxmox VE bare-metal to improve role separation and resilience:
HPE ProLiant DL20 Gen11
└── Proxmox VE
├── VM Windows Server 2025 (DC)
└── VM Windows Server 2025 (Applications)

Storage layout​

The server only provides HPE Embedded Intel VROC SATA Software RAID, so I plan to manage storage directly in Proxmox using ZFS RAID 1, structured as follows:
HPE ProLiant DL20 Gen11
└── Proxmox VE
├── ZFS RAID 1 – Pool 1 (OS)
│ ├── 2×480 GB
│ ├── ~200 GB → Windows Server 2025 VM (DC)
│ └── ~280 GB → Windows Server 2025 VM (Applications)
└── ZFS RAID 1 – Pool 2 (DATA)
└── 2×960 GB

Questions​

  • Does this approach make sense in this context?
  • Are there any known pitfalls when migrating from Windows bare-metal to Proxmox?
  • Best practices regarding ZFS, OS vs DATA pools, and boot disk layout?
I can provide detailed hardware specs, disk layout, or logs if needed.

Thank you in advance for your feedback. ♥️
 
  • Like
Reactions: ThoSo
It's not clear if you know the following page: https://pve.proxmox.com/wiki/Windows_2025_guest_best_practices - maybe some of those hints are helpful for you...
Thank you for the link and the references — they are definitely useful and I’ll review them.

At this stage, however, I’m mainly looking for feedback on the initial design choices, before going deeper into guest tuning:
  • moving from Windows Server bare-metal to Proxmox as a hypervisor
  • separating AD/DNS/DHCP from application workloads into distinct VMs
  • using ZFS RAID1 on local SSDs given the hardware constraints (VROC SATA only)
  • OS pool vs DATA pool layout

In other words, I’d like to validate whether this overall approach and storage layout make sense as a starting point, before fine-tuning Windows guest settings.

Any feedback or alternative recommendations on the architecture itself would be very welcome.
 
I know what forum this is, but out of curiosity, if you’re a Windows-only setup why not Hyper-V? With Standard you can have 2 VMs if the host is only Hyper-V.
That’s a fair question.

Hyper-V was indeed considered, especially since this is currently a Windows-centric environment.
However, my main goal is to avoid tightly coupling the hypervisor with the Windows / Active Directory stack again.

With Proxmox, I’m mainly looking for:
  • a minimal and stable hypervisor layer, independent from Windows patching, hardening and role changes
  • simple snapshot / rollback and bare-metal recovery workflows
  • a clearer separation between infrastructure services (AD / DNS / DHCP) and the virtualization host
  • more flexibility on the storage side (especially with ZFS)
I also have a second server available. In case of an issue, the idea would be to temporarily move VM images or snapshots to the other host to keep services running, with the possibility to evolve towards some form of HA in the future.

In addition, I already use Active Image Protector, which is compatible with Proxmox, and fits well into this backup and recovery strategy.

The goal is not to run many different OS types, but to reduce complexity and blast radius after the issues experienced with Windows Server bare-metal.

That said, I’m open to feedback if you think Hyper-V would be a better fit given these constraints.
 
P2V for Windows can be done, sure. We moved a few Windows servers from one KVM to another so it was a bit easier, but we still needed to add the SCSI driver to boot using SCSI, and the network card driver/config. And of course it can't be an OEM license if you're changing hardware. There are various threads discussing not using a "host" CPU type in the VM, or other workarounds for nested virtualization especially on more recent versions of Server.

Technically on Hyper-V the host can be a standalone server and not joined to AD...in fact I'd recommend that if the DC is a VM.