Proxmox VE is out, updating the underlying OS to Debian 13, and the underlying stack is getting updated as well. With the new release, we get features like RAIDZ expansion, a new fabric for the SDN stack feature, the ability to snapshot thick-provisioned volumes stored on iSCSI and Fibre Channel storage, high-availability affinity, and more.
Proxmox VE 9 is Out with Big New Features
If you just want to get Proxmox VE 9, you can either download and install via the ISO as usual, or if you are upgrading from PVE 8, then you usually want to update all packages, and then run the pve8to9 tool. On the Tyan GC70-B8033 1U AMD EPYC Milan server we reviewed we got a yellow amd-microcode warning. It was a small bummer that the amd-microcode tool did not install as suggested with apt install amd-microcode.
Once you do that, if you have a simple no-subscription test node, then it is three commands to get the system updated. If you have Ceph running, a clustered setup, or are running an enterprise subscription, it is certainly a case where you want to be much more careful. Check out the official upgrade documentation here.
On the ZFS side, the new version is ZFS 2.3.3. That is a big deal since ZFS 2.3.0 was where RAIDZ expansion was added, meaning we get that feature in this release. STH has been showing ZFS on Proxmox VE since PVE 3.4 in 2015, and a decade ago, something like this was a dream feature. We hope the Promxox team continues to develop the ZFS storage admin UI.
We also found that the ZFS ARC memory consumption is now being reported in the resources. Here is a quick look at the ZFS ARC being added to the resources memory tab after a PVE 8 to 9 upgrade.
The new HA rules is a neat feature. It helps manage where high-availability resources get placed. For example, you can specify which nodes to place HA resources on.
Additionally, you can specify that certain groups of resources should be placed on the same node or separate nodes.
On the networking side, there is a new fabrics feature allowing you to setup OpenFabric or OSPF fabrics. One of the new features that will fall under the radar, but it is likely some of our readers have sen is a new pve-network-interface-pinning tool.
This tool helps bind a MAC address and an interface name so you do not have an issue where you upgrade pve8to9, reboot, and find the interface name has changed and are locked out of a system. If you have ever had that happen, you will be very excited by this little tool.
Final Words
For a lot of us, a new Proxmox VE version is a lot of work. At the same time, when we go back to the archives and look at how far we have come in the last few years, it is always great to see new features come aboard the platform.
If you want to see all of the new features, you can find the full list here.
Debian 13? Did they get the release version a couple of days early or did they base their build on a pre-release version?
It’s a little convoluted, but the Debian version is frozen, and the Proxmox guys were apparently done. Technically, there’s no reason not to release it, other than the fringe case of somebody starting proxmox for the first time.
Hi, how’s it going? :)
The helper scripts aren’t set up for 9, either. All props to the developers, just your occasional reminder why the bleeding edge is just that.
My little test server just upgraded itself to 9 without issues. I didn’t bother shutting down my containers first either, and yolo’d the upgrade from a noVNC browser popup. Everything kept on trucking until reboot. The web UI did fail to load new pages during the install, but that seems forgivable.
I had vaguely hoped I might get a fresher kernel with this and get rid of some pesky N150 iGPU driver crashes, but today was not that day. 6.14.8-2 remains the latest kernel available on either side of the upgrade.
LOL, still has the NVidia frame buffer hang at install that has been there since what, 2020? They black list it in the kernel after install, but not before. Nothing like kicking the can forward.