Proxmox VE 9.2 with Dynamic Load Balancer and Linux Kernel 7.0

Vienna-based Proxmox Server Solutions GmbH has released its Proxmox Virtual Environment (VE) 9.2. It is based on Debian GNU/Linux 13.5 “Trixie”, including updated packages and bug fixes. Proxmox VE 9.2 uses a customized Linux kernel 7.0. QEMU has been updated to version 11.0, providing emulation and virtualization for virtual machines (VMs). For Linux containers, LXC version 7.0 is used. The storage stack is formed by OpenZFS 2.4, while distributed storage solutions are provided by Ceph 19.2.3 “Squid” or Ceph 20.2.1 “Tentacle”.

The Proxmox team considers the new Dynamic Load Balancer to be a central new feature. In dynamic scheduling mode, the Cluster Resource Scheduler (CRS) uses real-time metrics on resource utilization of nodes and guests to evenly distribute HA-managed guests across the cluster. The new load balancer can automatically migrate HA-managed guests to reduce imbalances in cluster node utilization. The behavior and sensitivity of the load balancer can be adjusted via various parameters in the HA section or in the datacenter options.

Improvements to the Software-Defined Networking (SDN) Stack

Proxmox VE 9.2 expands its capabilities to support modern network architectures. New additions include WireGuard and BGP as fabric protocols. Furthermore, Proxmox VE now supports route maps and prefix lists for fine-grained filtering of BGP and EVPN routes. Also new are the redistribution of routes in OSPF fabrics, additional options for configuring EVPN controllers, and support for IPv6 underlays in EVPN environments.

Custom CPU models are intended to allow administrators to define tailored virtual processors for specific workloads while standardizing the CPUs used within the cluster. This can help avoid compatibility issues within the cluster, among other benefits. A new section under “Datacenter / Guest Resources/Hardware” allows for the creation, editing, and removal of custom CPU models directly in the web interface. The selection area for CPU flags also displays the functions supported on the individual cluster nodes.

HA Deactivatable for Cluster Maintenance

During maintenance work in the Proxmox cluster, HA services can sometimes be more of a hindrance than a help. Using the new CRM commands “disarm-ha” and “arm-ha”, the HA stack can be deactivated cluster-wide in one go or reactivated later. This allows planned maintenance work on the entire cluster, including changes to the cluster network configuration, to be carried out without triggering fencing mechanisms for individual nodes, for example. The state of HA resources is preserved during deactivation, so resources automatically return to their previous state after maintenance is completed.

All improvements and new features, as well as potential issues when upgrading from Proxmox VE 9.1 to 9.2, are described in detail in the Proxmox Roadmap. Proxmox VE 9.2 is available as open-source software immediately for download and can be used free of charge. Prices for optional support subscriptions have been moderately increased by a good four percent. Access to the Enterprise repository now costs from 120 Euros net per year. Professional support costs between 370 and 1,100 Euros net per year per CPU socket, depending on the package.

(vbr)