Optiplex Proxmox Cluster

So I received 3 Optiplex 3050s from work.
16 GB 2400 RAM
128 GB intel SATA drive

I currently have a decent homelab running proxmox 7.4 and I wanna make a decent “Production” server for some of the services I like while also trying to migrate to Proxmox 8.1. I need help in decided what hardware I should go with considering the boards age and my desired services.

Services I want to run:

  • Jellyfin
  • Git repository
  • PiHole
  • Remote Gaming PC
  • Folding@Home or BOINC instance
  • Minecraft server

Things I think I need:

  • LP BPO GPU for compute and video transcoding
  • NIC for Clustering
  • Dedicated Cluster switch
  • RAM and Storage upgrades
  • potentially a second NAS

I don’t quite know what I need to have to do all these things, but I do have somewhat of a budget. I was looking at the Arc 310, RX 6400, and Quadro T400 as LP BPO single slot GPUs to put in them. I’m not sure what the best one for F@H would be (I like the idea, but I doubt the 310 can do any of this). I think the NVIDIA card would do transcoding the best.

I think I can figure everything else out, but I’m very undereducated on GPUs for none gaming and Clustering. Should I even try running them as a cluster or just 3 separate nodes? And do I need to upgrade the gigabit networking they all have built in? I’m just a little overwhelmed on what should I do next?

Like anything, aim for simplicity and plan to upgrade in the future. Gigabit is honestly plenty for most people, just ensure you use vlans so can deploy multiple networks as needed. Upgrading to 10G in the future isn’t too hard, especially with VMs that can be moved.

Anything that uses lots of storage (like jellyfin) will essentially be pinned to the NAS with the storage, unless you use NFS/SMB. Same goes for using GPUs, it basically pings the VM to that physical host. So I’d plan for some to be semi-permanent, and some to be movable. With the right storage config (replication for example) you can redeploy those semi-permanent VMs on another proxmox node if one goes offline.

With proxmox you really lose nothing by running them as a cluster, so I’d go with that. The only gotchas is that you can’t start one machine by itself, and that you should follow the node removal procedure properly when decommissioning.