Upgrade current v8 cluster to v9 by adding two new nodes?

[LSITGUY]

Hello all,

I'm hoping someone can guide me.

I would like to upgrade my current Proxmox cluster running v8 to v9 and I thought that adding two "new" nodes to the cluster might be a good way to do that.

Here's a quick summary of my setup and what I plan to achieve:

• I have a 5-node PVE cluster with shared storage (NFS only). All five nodes are running pve version 8.3.4
• I have a couple of servers that used to have oVirt on them. All VMs have been moved from the oVirt cluster to Proxmox.
• I would like to re-purpose those servers by adding them to the current Proxmox cluster.
• These "new" nodes are not of the same make/model as the current Proxmox nodes.

I was thinking of doing the following:

• Install pve version 9 on the "new" nodes
• Add them to the current pve cluster
• Pick one of the v8.3.4 servers and:
  • Backup the VMs (just in case)
  • Backup the /etc directory of the node (and maybe other files)
  • Migrate VMs to one of the v9 nodes
• Completely remove the v8 node from the cluster
• Install v9 from ISO on said node
• Re-add to the cluster
• Migrate VMs back to their node.
• Follow the same steps for the other four v8.3.4 nodes.

Will this work? Or am I completely out-to-lunch here?

thanks in advance for your help
Guy

 

[LnxBil]

Will this work? Or am I completely out-to-lunch here?

Yes, that will most probably work. Why not just upgrade the nodes to the newer version?

Caveats:

• Live migration from and to non-identical CPUs can result in problems (hangs, crashes, etc.) best to do an offline migration.
• Removing and adding nodes with the same name could be tricky. Stick to the documentation.

However, I would build a nested PVE cluster that tries to match your current cluster, just with smaller disk and memory and try everything out there. It does only cost time and will give you the feeling how it would go and you have hands-on experience.

 

[LSITGUY]

Yes, that will most probably work. Why not just upgrade the nodes to the newer version?

Caveats:

• Live migration from and to non-identical CPUs can result in problems (hangs, crashes, etc.) best to do an offline migration.
• Removing and adding nodes with the same name could be tricky. Stick to the documentation.

However, I would build a nested PVE cluster that tries to match your current cluster, just with smaller disk and memory and try everything out there. It does only cost time and will give you the feeling how it would go and you have hands-on experience.

Ok, thank you for your time. I will keep you comments in mind. Very useful. thanks again.