Proxmox VE 9.0 is now available, bringing a range of practical, under-the-hood improvements that continue to push the platform forward. If you’re running clusters at scale, managing sensitive workloads, or planning a move away from VMware, this release deserves your attention.
Built on Debian 13 (“Trixie”) and shipping with Linux kernel 6.14, QEMU 10, and ZFS 2.3.3, Proxmox VE 9.0 introduces enhancements across HA, SDN, storage, container behaviour, and management workflows. It’s a release focused on reliability, capability, and deployment flexibility.
If you’re planning a Proxmox 8 to 9 upgrade, this post covers the most important changes and what to expect. It also highlights where these features make the biggest difference in the field.
Let’s step through some of the key changes!
Snapshots on Shared LVM Storage
Proxmox VE 9.0 introduces long-awaited support for snapshots on thick-provisioned LVM shared storage, such as iSCSI or Fibre Channel-based SANs. This is achieved by implementing snapshots as volume chains, where a new volume records only the differences from its parent snapshot volume. It’s a powerful, storage-independent solution that brings true snapshot capability to shared block storage without needing alternative storage solutions or complex workarounds.
Historically, if you wanted to take VM snapshots on Proxmox and you weren’t using ZFS, you were out of options. Shared LVM-backed storage didn’t support snapshots, and while qcow2 technically supported them, it wasn’t suitable for most real-world deployments. qcow2 is a file-based format that requires a traditional filesystem underneath, like ext4 or NFS, and isn’t usable on raw LVM volumes or SAN LUNs. Even on local storage, qcow2 introduces performance overhead and operational friction. Most production environments avoided it for good reason.
This new snapshot model is ideal for customers with large-scale SAN infrastructure who want to preserve their existing investment. Previously, lack of snapshot support on shared LVM volumes was a dealbreaker for many Fibre Channel or iSCSI users. This release changes that by enabling consistent, hardware-independent snapshots without giving up performance or compatibility. It also benefits those coming from clustered filesystem environments like VMFS or GFS2, who now have a Proxmox-native solution that doesn’t require third-party storage plugins.
High-Availability Affinity Rules
Proxmox VE 9 introduces a more intelligent High Availability system through node and resource affinity rules. These rules give administrators the ability to define exactly how HA-enabled resources (like VMs or containers) are placed within a cluster.
You can now:
- Assign specific nodes as preferred hosts for particular resources
- Keep dependent workloads together to minimise latency
- Spread redundant workloads across different nodes to improve fault tolerance
This is especially useful in multi-tier application environments. For example, a redundant application deployed across multiple nodes can now explicitly avoid resource collisions, ensuring that one node failure doesn’t take down all instances. These affinity rules bring Proxmox’s HA capabilities more in line with what you’d expect from enterprise-grade cluster managers.
Proxmox SDN Fabrics and OSPF Support
This release enhances the Proxmox SDN stack with support for “Fabrics”, a major step forward in simplifying and scaling dynamically routed infrastructure. The new SDN Fabrics feature supports OpenFabric and OSPF protocols, allowing Proxmox nodes to form resilient, automatically routed topologies. This enables use cases such as Ethernet VPN (EVPN) underlays, full-mesh Ceph networks, or multi-site cluster routing without static network configuration.
SDN Fabrics support multiple NICs and automatic failover, making it possible to build two-tier spine-leaf network architectures that improve both redundancy and throughput.
This is a foundational feature for sites deploying Proxmox across multiple datacentres or rack domains, or where routing flexibility is needed. With fabrics, we can build routed topologies with dynamic path awareness – and even use them as underlays for encrypted VPN mesh networks or site-to-site overlays.
An encrypted VPN mesh network connects multiple Proxmox nodes or sites using secure, encrypted tunnels that automatically route traffic between peers. Instead of relying on a central hub, each node can communicate directly with any other node, forming a resilient mesh. This is particularly useful for linking distributed environments where security and fault tolerance are critical. By using SDN fabrics as the underlay, the routing is handled dynamically, simplifying the overlay setup and making the mesh more adaptive to changes in topology or network state.
We’re planning a larger blog post (or maybe a series!) to explain this in depth.
A Few Cautions Before You Upgrade
At the time of writing, Debian 13 (“Trixie”) has not yet reached general availability. The upstream GA release is expected within days, but Proxmox VE 9 is built against the Debian 13 snapshot from May. While this is unlikely to cause issues, users should be aware that the underlying platform is technically pre-GA.
In addition, Proxmox Backup Server 3 is not compatible with Proxmox VE 9.0. Support will resume with Proxmox Backup Server 4, which is expected to be released shortly which has just been released!
Proxmox Backup Server 4 has just been released with full compatibility for Proxmox VE 8.x and 9.x. Proxmox VE 8x also supports backing up to PBS 4, so no worries there.
Final Thoughts
Proxmox VE 9.0 doesn’t come with your typical flashy headline features, but that’s part of what makes it good. If you’re wondering what’s new in Proxmox 9, the answer is a long list of thoughtful, production-focused improvements. This release focuses on practical improvements that make life easier in real environments: shared storage snapshots, affinity-based HA, SDN fabrics that actually reduce complexity, and live migration that better respects network state.
If you’re already using Proxmox, v9 is a strong upgrade. If you’re on VMware and looking for a clean exit, this release makes Proxmox even more viable as a replacement platform. It respects your existing infrastructure investments, and it keeps improving without locking you into someone else’s commission incentives.
Need help upgrading, migrating off VMware,
or getting Proxmox running properly in production?
