Proxmox VE 9 is not a minor update. It is the platform's first major release since Broadcom's acquisition of VMware turned the virtualization market upside down, and it arrives at the exact moment when thousands of companies are deciding which hypervisor to build the next decade of their infrastructure on. If you are weighing your options, our Proxmox vs VMware comparison covers that context in detail.
The answer from Proxmox Virtual Environment to that opportunity is emphatic: a technology stack renewed from top to bottom (Debian 13, kernel 6.14, QEMU 10, LXC 6, ZFS 2.3) and, on top of it, features that directly address the gaps that separated Proxmox from the most demanding enterprise environments: snapshots on shared LVM storage, SDN networks with dynamic routing protocols and affinity rules in the high availability manager.
In this article we go through the most relevant new features in Proxmox VE 9, explain why they matter in real production environments, detail the upgrade path from PVE 8 and explain what this release means for customers of EasyDataHost's cloud platform.
A New Technical Foundation: Debian 13, Kernel 6.14, QEMU 10 and LXC 6
Proxmox VE 9 is built on Debian 13 "Trixie", guaranteeing years of security support on a stable, audited base. The base change brings modern versions of the whole stack: glibc, systemd, OpenSSL and the networking tools are all refreshed at once, something that in a hypervisor has a direct impact on hardware compatibility and performance.
The Linux 6.14 kernel is the most visible part of that renewal: support for the latest AMD and Intel CPU generations, improvements in KVM (the hypervisor Proxmox is built on), a more efficient scheduler on many-core systems and better support for NVMe controllers and 25/100 GbE network cards. On recently purchased servers this translates into fewer driver headaches and more performance per watt.
QEMU 10 brings emulation improvements and faster, more robust live migration, with shorter pause times when moving virtual machines between nodes. LXC 6 modernises the system container layer, with better isolation and compatibility with the newest distributions inside containers. Taken together, the virtualization stack is now aligned with the latest in the open source ecosystem.
Storage: ZFS 2.3 with RAIDZ Expansion and Snapshots for Shared LVM
The headline storage feature is ZFS 2.3 with RAIDZ expansion. Until now, growing a RAIDZ vdev was impossible without destroying the pool: if you had a 6-disk RAIDZ2 and wanted to go to 8, you had to build a new pool and migrate all the data. With ZFS 2.3 you can add disks to an existing RAIDZ, one at a time and with the pool in production. For servers with free drive bays this completely changes capacity planning: you can start small and grow with actual demand.
The second improvement tackles a long-standing pain point in SAN-backed environments: Proxmox VE 9 introduces snapshot support for thick-provisioned LVM volumes on shared iSCSI or Fibre Channel storage. Until this release, anyone connecting Proxmox to a SAN over iSCSI/FC with LVM had to give up VM snapshots — a limitation that weighed heavily in migrations from VMware, where snapshots on shared datastores are taken for granted. With PVE 9 that scenario is covered through chained snapshot volumes, and one of the most cited technical barriers to migrating disappears.
Both improvements point in the same direction: storage should no longer dictate your architecture. Whether you run local ZFS on NVMe disks or connect a corporate storage array over Fibre Channel, Proxmox VE 9 now offers snapshots, cloning and a consistent day-to-day operation.
SDN Fabrics: OpenFabric and OSPF for Complex Networks
The Proxmox SDN (Software-Defined Networking) stack takes a big step forward with fabrics: routed networks between nodes configured from the web interface itself, using OpenFabric or OSPF as the dynamic routing protocol. Instead of hand-maintaining static route configurations between nodes, the fabric discovers the topology and automatically recalculates paths when a link goes down.
Where does this shine? In two very specific scenarios: full-mesh networks for Ceph, where every node connects directly to the others without needing high-capacity switches, and clusters spread across several rooms or racks with multiple redundant paths. What used to require advanced FRRouting knowledge and manual configuration files is now defined in a few clicks and managed (and versioned) by Proxmox itself.
High Availability with Affinity Rules and Interface Improvements
The high availability manager finally gains affinity and anti-affinity rules. With them you can specify that a group of VMs must run together on the same node (affinity, for instance to minimise latency between an application and its database) or that they must never share the same host (anti-affinity, for instance the two nodes of a replicated database cluster). When HA relocates machines after a node failure, it honours these rules automatically.
Anyone coming from VMware will recognise the concept: these are the classic DRS affinity rules, one of the features most sorely missed in Proxmox. Their arrival removes another of the traditional arguments against migrating.
On the interface front, Proxmox VE 9 debuts a revamped mobile interface, built on the same framework as Proxmox Datacenter Manager. Checking cluster status, starting or stopping a VM or reviewing a task from your phone no longer means wrestling with a shrunken desktop web page: routine on-call operations can be handled from a mobile device.
Comparison Table: Proxmox VE 8 vs Proxmox VE 9
The following table summarises the key differences between the two major releases:
| Component | Proxmox VE 8 | Proxmox VE 9 |
|---|---|---|
| Base system | Debian 12 "Bookworm" | Debian 13 "Trixie" |
| Linux kernel | 6.8 | 6.14 |
| QEMU | 9.x | 10 |
| LXC | 5.x | 6 |
| ZFS | 2.2 | 2.3 (RAIDZ expansion) |
| Snapshots on shared thick LVM (iSCSI/FC) | No | Yes |
| SDN fabrics (OpenFabric, OSPF) | No | Yes |
| HA affinity/anti-affinity rules | No | Yes |
| Mobile interface | Basic | Revamped |
Upgrade Path from Proxmox VE 8
Proxmox supports the in-place upgrade from PVE 8 to PVE 9 without reinstalling. The process will be familiar to anyone who already went from 7 to 8, but since it involves a major Debian version jump it should be executed methodically:
- check_circle Start from the latest PVE 8.x: bring the node up to the latest packages of the 8 branch before starting the major version jump.
-
check_circle
Run
pve8to9: the official checker tool detects incompatibilities (obsolete packages, legacy network configurations, unsupported storage) before anything is touched. Re-run it until no critical warnings remain. -
check_circle
Take full backups before upgrading: verified backups of every VM and container, plus
/etc/pve. A major operating system upgrade should never run without a rollback plan. - check_circle Update the repositories: switch the APT sources from Bookworm to Trixie (both Debian and Proxmox repositories) and run the full system upgrade.
- check_circle In a cluster, go node by node: live-migrate the workloads off the node being upgraded, upgrade it, verify it rejoins the quorum and move on to the next one. Never upgrade all nodes at once.
Practical recommendation:
Do not upgrade production in the first week. Rehearse the jump on a lab node or a secondary host with non-critical workloads first, run pve8to9 --full on every node and schedule the maintenance window with a verified backup and a written rollback procedure. In HA clusters, temporarily disable the HA services on the node under maintenance before upgrading it.
With these precautions the upgrade is a predictable process: VMs keep their configuration, storage is recognised without changes and the cluster stays in service throughout, thanks to live migration.
What It Means for EasyDataHost Customers
At EasyDataHost we run our cloud platform on Proxmox 9.x, in our own datacenter in Spain with ISO 27001 certification and ENS compliance. For our customers, the new features in this release translate into direct benefits with no action required on their side:
- arrow_right Faster live migrations thanks to QEMU 10, with maintenance windows that are practically imperceptible.
- arrow_right Greater resilience: anti-affinity rules guarantee that a customer's redundant VMs never end up on the same physical node.
- arrow_right Storage growth without interruptions thanks to ZFS 2.3 RAIDZ expansion on the storage nodes.
- arrow_right A base supported for years: Debian 13 and kernel 6.14 secure long-term security patches and hardware compatibility.
And if you are still on VMware, Proxmox VE 9 is the best entry point to date: snapshots on shared storage and affinity rules remove the two most common technical objections. Our guide on how to migrate from VMware to Proxmox explains the process step by step, and if you would rather assess it together, contact our team for a no-obligation analysis.
Frequently Asked Questions
Can I upgrade directly from Proxmox VE 8 to Proxmox VE 9?
Yes, the in-place upgrade is supported from the latest PVE 8.x release. Run the pve8to9 checker first to detect incompatibilities, take full backups of VMs and containers, update the repositories to Trixie and, in clusters, proceed node by node while live-migrating workloads.
What does ZFS 2.3 RAIDZ expansion bring?
It lets you add disks to an existing RAIDZ vdev without destroying the pool or migrating data. Until now, growing a RAIDZ meant building a new pool and copying everything over. With ZFS 2.3, growth is incremental, disk by disk, with the pool in production.
Is Proxmox VE 9 a real alternative to VMware?
For the vast majority of workloads, yes. PVE 9 covers high availability with affinity rules, snapshots on shared iSCSI/FC storage, advanced SDN, integrated backup and a complete API, with no per-core licensing or mandatory subscriptions. After Broadcom's licensing changes, migrating to Proxmox has become the most common route.
Conclusion
Proxmox VE 9 is the project's most ambitious release to date, and it arrives at just the right time: exactly when the market is looking for a solid alternative to VMware. The key takeaways:
- arrow_right Renewed foundation: Debian 13 "Trixie", kernel 6.14, QEMU 10 and LXC 6 guarantee modern hardware compatibility and long-term support.
- arrow_right Storage without strings attached: RAIDZ expansion in ZFS 2.3 and snapshots on thick LVM over iSCSI/FC, the historical gap that held back VMware migrations.
- arrow_right Enterprise operations: SDN fabrics with OpenFabric/OSPF, affinity and anti-affinity rules in HA and a revamped mobile interface.
-
arrow_right
Predictable upgrade: in-place from PVE 8 with
pve8to9, prior backups and node-by-node upgrades in clusters.