Skip to content
A Proxmox VE node running Proxmox's own Ceph monitor and manager beside croit FlashStore OSDs on eight Optane drives

FlashStore OSD on Proxmox VE: What It Took on One Node

croit’s FlashStore OSD is a Rust replacement for ceph-osd and BlueStore, in public preview since 9th November 2026. This deploys it hyperconverged on a single Proxmox VE 9.2 node: Ceph Umbrella 21.2.0 from the PVE wizard for the monitor and manager, FlashStore OSDs on Intel Optane P4800X drives. It covers why Umbrella is the only Proxmox release that matches, formatting the drives to 4 KiB and the controller reset that aborts a format, what croit-flashstore-osd –check reports, why the shipped unit cannot read Proxmox’s ceph.conf and the drop-in that fixes it, the NVMe passthrough Proxmox’s kernel refuses and the direct backend that works, preparing OSDs by PCI address, what the PVE OSD panel makes of them, a pool, RBD storage and a Fisher-Price OS (Windows) VM installed onto it, CrystalDiskMark from inside that VM read against the drives’ own counters, Ceph’s memory with all eight OSDs active, three VMs at once on a 1 GiB memory target, and a setup script that does all of it, tested by rebuilding an OSD from nothing.

11th November 2026 · 35 min · 7721 words · Damien Dye
Three customer sites feeding one Caddy, which hands each its own token, in front of one Proxmox Datacenter Manager

One PXE Estate for Every Customer: Unattended Installs with Proxmox Datacenter Manager

Proxmox Datacenter Manager 1.1 can serve answer files to the Proxmox automated installer, which makes it a natural install server for an MSP looking after many customer sites. The use case is a customer who wanted every site built with its own naming, root password and SSH keys, by plugging a server into the right VLAN and booting it, with no serial numbers to collect and no answer file per machine; this post covers that booting and installing half, and leaves joining the cluster aside. It cannot match an answer on the site, and it cannot give a node a first-boot script, and both are shown from PDM’s own source. This builds one PXE estate for every site anyway: one boot image carrying a non-secret catch token, DNS TXT discovery so the image holds no URL, and a Caddy proxy that swaps the catch token for the site’s own bearer token by source subnet. It covers how PDM picks an answer, prepare-iso for iPXE, the DNS records, dnsmasq and the iPXE loop, a hardened Caddy that serves the whole PDM UI on 443 with 8443 closed by nftables, logging which token was used by name only, the post-installation webhook, adding Proxmox Backup Server, and the trade-offs that remain.

7th November 2026 · 35 min · 8070 words · Damien Dye
Guest writes on the left, VDO in the middle, and only about a quarter of the bytes reaching the drive on the right

VDO Under Proxmox Local LVM: Fewer Bytes Reach the Drive

VDO is the kernel’s deduplication and compression layer. Under a thick LVM volume group it gives Proxmox VE local disks that write one copy of every block the VMs share, so far fewer bytes ever reach the drive. This builds it thick on thick on a Proxmox VE 9 host, with VDO sized to the space it really has so nothing above it can overcommit, the lvm.conf scan_lvs setting and its filter, and VDO’s threads matched to the CPU count. It measures what deduplication saved on a template and six full clones, two ISOs, zeros and random data, how many bytes the drive actually received against what the guest wrote, and what VDO costs against the same Intel Optane with no VDO at all. It then builds the pool with VDO’s index and journals on NVMe and the data on a second device, proves the layout with vdolistmetadata, measures how the I/O splits, then puts VDO on top of bcache with one NVMe partitioned between the bcache cache and VDO’s metadata, and measures how many bytes reach the cache with and without VDO. It reasons through what all of this means on HDD hardware RAID6, and covers discard on delete.

27th October 2026 · 28 min · 6177 words · Damien Dye
Two Proxmox defaults: Hyper-V enlightenments switched on by the VM's OS Type, and CPU microcode switched on by how the host was installed

Two Proxmox VE 9 Defaults You Might Not Have: Hyper-V Enlightenments and CPU Microcode

Proxmox VE 9 turns on Hyper-V enlightenments for Fisher-Price OS (Windows) guests and installs CPU microcode on new hosts, but only if the VM was given a Microsoft Windows OS type and the host came from the 9.x installer. This covers what the enlightenments do, the exact rules qemu-server uses to pick them (measured on a live 9.2 host), why machine version pinning keeps older VMs off the July 2026 hv-evmcs fix, the pve-qemu-kvm build that fix needs and what each QEMU release since 10.0 fixed for those guests, why they belong on q35 and never i440fx (which the wizard still hands most of them), the two flags you can set by hand, how to prove what a VM is running, where microcode comes from (board firmware, Debian and Proxmox’s own repository), why an upgraded host has none (with an Intel node four Intel releases behind as the example), the initramfs rebuild that installing it triggers and the older kernels that rebuild misses, the Zen 5 RDSEED failure that follows, enabling it, a rolling reboot across a Ceph cluster, and rolling back a bad update.

27th October 2026 · 25 min · 6016 words · Damien Dye
A licence key moving from a PC's firmware into a single virtual machine

Moving an OEM Licence Into a Proxmox VM, and What Moves With It

An OEM licence for the Fisher-Price OS (Windows) lives in the PC’s firmware as an ACPI table called MSDM, and QEMU will pass any such table to a guest without checking it. This covers what the table holds, a validator that refuses malformed tables before QEMU silently repairs their headers, storing them in the private half of the Proxmox cluster filesystem, Microsoft’s terms on moving an OEM licence, the one case they plainly allow, taking the table straight from the host when the licensed machine is the hypervisor, and why MAK, KMS, Active Directory and AVMA activation never touch the table.

4th October 2026 · 11 min · 2595 words · Damien Dye
A virtual machine's boot screen with the hypervisor's logo replaced by a product's own

Branding a Proxmox VM, and Keeping It Branded Through the Next Upgrade

A stock Proxmox VM shows Proxmox’s logo at boot, says Proxmox in its firmware vendor string and calls itself QEMU in every SMBIOS table. This replaces all of it with your product’s name: SMBIOS types 0, 1, 2, 3 and 11 with the serial number and SKU filled from the VM’s own name and size, the SeaBIOS splash and its colour trap, a branded OVMF built from Proxmox’s own tree or an option ROM driver that replaces the logo and the BGRT under Secure Boot without touching the firmware, and the web interface logo. Every change is placed so an upgrade cannot quietly undo it, and the one an upgrade can leave dangerously stale, the firmware, gets a hook that says so.

4th October 2026 · 24 min · 5661 words · Damien Dye
A browser connecting through Cloudflare Access to Windows VMs on a Proxmox host with Intel Arc Pro GPU virtual functions

Zero Trust VDI Without the Cloud Bill — Proxmox, Intel Arc Pro and Cloudflare Access

Part one: the architecture and the cost argument for replacing Azure Virtual Desktop with Proxmox, Intel Arc Pro SR-IOV and Cloudflare Access. On-prem OAuth for identity, browser-rendered RDP, a firewalled LXC tunnel on its own VLAN, KSM for memory density, and profiles on Ceph. The next post builds it.

2nd September 2026 · 17 min · 3677 words · Damien Dye

Who Actually Writes & Uses Ceph

I counted every commit to Ceph’s main branch for the last ten years — 69,613 of them from 1,718 people across 478 organisations — traced all 38 Steering Committee members to their employers, and gathered the publicly stated deployments. The result is a project that absorbed the loss of its founder and its second-largest contributor without dropping a beat, and is still shipping at the same rate today.

27th August 2026 · 36 min · 7504 words · Damien Dye
One NFS export mounted by both vCenter and Proxmox, so a VMDK crosses the network once instead of twice

VMware to Proxmox with Ansible — Build the Shells Before You Move a Byte

Reading a vSphere estate with the dynamic inventory, mirroring its VLANs into Proxmox SDN, and rebuilding every VM as a diskless shell before a single disk moves. Why every VMware info module hides the one field the migration needs, why vlan_id is three different types, and why one NFS export mounted twice turns the cutover into a local read.

26th August 2026 · 20 min · 4257 words · Damien Dye
An Ansible control node calling the Proxmox API, with the VM being built having no connection back

Creating Proxmox VMs with Ansible — The Host You Are Building Does Not Exist Yet

The community.proxmox collection is an API client, not a configuration agent, and that changes the shape of every playbook that uses it. Where the tasks actually run, why proxmox_kvm declines to converge rather than updating, why proxmox_disk and proxmox_nic are where disk and NIC changes belong, and the undocumented return value you will end up needing.

12th August 2026 · 25 min · 5224 words · Damien Dye