Weird issue in upgrading from 8.4 to the current 9.2.10

choctaw

New Member
May 14, 2026
1
0
1
Did something change recently?

I've been tasked with upgrading our clusters. I hadn't done the 8 to 9 upgrade in about year, and that was on my home lab, which somehow made me 'the expert' :rolleyes:. For my current position, I built a test cluster, ceph, etc. I did the upgrade from 8to9, zero problems as expected. So in the last 2 weeks, I was going to rebuild the cluster starting at 8.4 and document the entire procedure for others who will be helping upgrade the other clusters I don't manage.

I was surprised when I followed my own notes(which borrow from the official install instructions) to find that I could not get any of my test nodes in the cluster to upgrade to 8.4.19, which in turn means there is no pve8to9 script installed. Yes, the repos were changed to what is required, non-subscription for the time being. Any subsequent apt update/upgrade resulted in packages being held back, I even went as far as clearing out /var/lib/apt/lists to purge any possibly weird metadata. I forgot to mention that first update, which should have brought me up to 8.4.19 but stayed at 8.4, knocked out the gui, and the normal fixes for that didn't work(restarting pveproxy, pvedaemon, and running a pvecm updatecerts --force), there is no firewall, /etc/hosts is fine and so is /etc/network/interfaces.

A full version update did upgrade the node, but this seems to me, to fall completely out of line with the official procedure. This, as expected, fixed the temporary issue of the gui not functioning. So, I am becoming a little concerned about moving forward on this until I understand what the heck is or maybe isn't transpiring. And of course, I have 77 nodes they want upgraded before the end of the month(end of 8.4 support), and yes I know this place waited till the last minute to do this, but I've only been here a short time and it was dumped on me.

I'd appreciate any words of clarity, or some direction. I did look, but I couldn't find anything related to what I have going on.
thanks,
 
Hi @choctaw, welcome to the forum.

Did you save any command-line output from the time you were attempting the upgrade?

Frankly, based on the iterations of troubleshooting you described, but without the corresponding command output, it is impossible to provide a meaningful recommendation. There are simply too many possible causes, and the fact that a subsequent full upgrade worked does not tell us what prevented the earlier upgrade from proceeding as expected.

My recommendation would be to replicate your production environment as closely as practical (2-3 hosts should be sufficient) and perform the upgrade there while capturing the complete output. Whether you can restore a few production hosts into an isolated lab, run them virtually, or install a small test environment from scratch is something you will need to determine.

Once you have a repeatable procedure on the test cluster, you can apply it to the production nodes with considerably more confidence.

Another option would be to engage a trusted PVE partner to analyze the existing systems and help develop an upgrade plan. Given the history you describe, I would also want to understand whether the existing clusters have accumulated package, repository, or configuration changes over time. That history could be highly relevant to what you are seeing now.

Cheers.


Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox