Proxmox VE 9.2: Dynamic Load Balancer and a Fresh Tech Stack
Table of Contents
On May 21, 2026, Proxmox Server Solutions released Proxmox Virtual Environment 9.2. If you run Proxmox in a homelab or in production, this release mainly delivers one thing: less manual cluster management and a noticeably more modern foundation under the hood.
Unlike a pure bugfix release, there is real substance here — from automatic load balancing and expanded Software-Defined Networking (SDN) to maintenance features you will notice in day-to-day operations.
What Changed Under the Hood#
Proxmox VE 9.2 is based on Debian 13.5 “Trixie” and makes Linux Kernel 7.0 the new stable default. Updated core components include:
| Component | Version |
|---|---|
| QEMU | 11.0 |
| LXC | 7.0 |
| ZFS | 2.4 |
| Ceph Squid | 19.2.3 |
| Ceph Tentacle | 20.2.1 (new default) |
Especially relevant for Ceph users: Tentacle 20.2.1 is now the default version for fresh Ceph deployments. Squid remains available as an option. If you are coming from older Ceph versions, follow the upgrade documentation carefully — it is a multi-step process.
Dynamic Load Balancer: The Cluster Balances Itself#
The most noticeable feature in 9.2 is the Dynamic Load Balancer. It builds on the Cluster Resource Scheduler (CRS) and works in a new dynamic mode.
Until now, load distribution in Proxmox clusters was mostly manual work: watch VMs, spot hotspots, migrate. The Dynamic Load Balancer takes over that for HA-managed guests automatically. Real-time metrics from nodes and VMs feed into every placement decision.
In practice, that means:
- The load balancer detects imbalances in the cluster
- It automatically migrates HA guests to less loaded nodes
- All custom HA rules remain in place
- Behavior and sensitivity can be fine-tuned via parameters in the HA panel
For homelabs with three or four nodes, this may still be nice-to-have. In production environments with dozens of VMs, it can save noticeable admin time — and prevents a single node from becoming a bottleneck while others sit mostly idle.
SDN: WireGuard, BGP, and Finer Filtering#
Software-Defined Networking features were significantly expanded in 9.2:
- WireGuard and BGP as new fabric protocols
- Route Maps and Prefix Lists for granular BGP/EVPN filtering
- Route redistribution for OSPF fabrics
- Additional configuration options for EVPN controllers
- IPv6 underlay for EVPN
If you have mostly ignored SDN in Proxmox because it felt like overkill for simple setups, these tools can now model more complex network topologies — without reaching for external SDN solutions.
Custom CPU Models Directly in the GUI#
Specialized workloads sometimes need specific CPU flags or models. In 9.2, Custom CPU Models can be created, edited, and deleted directly under Datacenter → Guest Resources/Hardware — no CLI hacks required.
Also practical: the CPU Flags Selector shows supported flags per cluster node. That helps catch compatibility issues early, before a live migration fails.
HA Arm/Disarm: Maintenance Without Surprises#
Anyone who has done maintenance on an HA cluster knows the problem: the HA manager reacts to planned node shutdowns with fencing or unwanted failovers.
With the new disarm-ha and arm-ha commands, you can temporarily disable the HA stack cluster-wide:
- Before maintenance: disarm HA
- Perform maintenance — no unwanted HA actions
- After maintenance: arm HA again — resources return to their previous state
It sounds simple, but in practice it saves a lot of stress during planned updates or hardware swaps.
Other Changes#
Beyond the big features, there are a few details that are useful in everyday use:
- Microsoft and Windows UEFI 2023 certificates can be imported via GUI and API
- Numerous bug fixes and performance improvements across the platform
Upgrade: How to Get to 9.2#
Good news: the upgrade is possible via apt — from both 9.1 and older 9.x versions.
| Starting point | Approach |
|---|---|
| Proxmox VE 9.1 | Direct upgrade via apt or GUI |
| Proxmox VE 8.4 | Upgrade using the 8 → 9 guide |
| Debian 13 Trixie | Installation on existing Debian is possible |
Proxmox recommendation: switch to [pve-enterprise] during the upgrade for the most stable experience.
Proxmox VE 8.4 still receives security updates until August 2026 — so there is time for a planned migration.
For Ceph upgrades: upgrade Ceph first (e.g. Reef → Squid on PVE 8.4), then Proxmox to 9.2, then optionally Ceph Squid → Tentacle. The order matters.
# Typical upgrade on a node (after backup!)
apt update
apt dist-upgrade
Of course, take a snapshot/backup first and ideally upgrade node by node in the cluster.
Conclusion#
Proxmox VE 9.2 is not a revolutionary release, but a very thoughtful one. The Dynamic Load Balancer and HA arm/disarm features address real admin pain points. The tech stack refresh with Kernel 7.0, QEMU 11, and ZFS 2.4 provides a solid foundation.
For homelab operators, the upgrade is worth it once your own dependencies (Ceph, custom scripts, backup jobs) have been checked. In production environments, validate in a test environment first — as always.