I’m new to the forum but I’ve been following the youtube channel for quite some time. Sorry for the long thread but I want to give context.
A month or so ago I retired my old desktop PC and decided to turn it into a homelab. After watching several videos and following many more guides I was able to stand up a Proxmox with a TrueNAS Scale VM and another Ubuntu VM which I pass through the GPU to use with Jellyfin and Photoprism containers.
I’m leaning towards having this PC dedicated to NAS on BM because this is turning into a hobby of mine and it’s the perfect excuse to build another PC/server for everything else I plan to do (pfSense, Pi-hole, Traefik, VaultWarden, NextCloud, Navidrome and so on). Some of these I already have working as LXCs.
My current build is:
i9-9600k
Asus Z370-A PRO
4x16GB DDR4-2400
2x SSD crucial BX500 1tb
2x HDD Seagate Barracuda 8tb
GTX 1070 Ti
I’m turning this build my NAS server because it’s a full tower with plenty of space for drives and the motherboard has several PCIe slots available. Maybe I can repurpose the crucial SSDs for the other server and use a couple of older SSDs for the boot drive here. I’m definitely getting an HBA PCIe card, adding a 3rd Barracuda drive and getting a 10gb NIC. What I’d like to know is what makes more sense, if keeping the GPU in this build and add Jellyfin, Photoprism and something else that might use it as Apps or have those in this other server I’m building (as containers in a VM like I have now or maybe even as unprivileged LXCs with GPU passthrough)? If the former, is 64GB RAM an overkill? I read ZFS will use all the RAM it can have but I don’t know if it makes a difference.
I plan to ask about recommendations for this other server and maybe even a workstation but I want to settle the NAS first.
Any help will be greatly appreciated.
Keep things simple. Multiple hypervisors and nesting is just adding complexity, causes trouble and is just more work. Some LXC containers + TrueNAS hypervisor with k3s+docker containers + nested Ubuntu with other containers and Passthrough? that’s just a mess.
Then just do storage and delegate all things virtualization to another machine. Way easier and a bare-metal storage server (NAS) is worth a lot in many aspects. One server for storage, the other one for
There is no such thing as overkill when it comes to memory capacity. more = better
Well, if the NAS has memory, networking and HBA/SATA for storage. The other one just needs as much CPU and memory as VMs demand. And preferably two x8 slots for GPU and NIC in case you need a GPU and beefier networking.
Thanks for the reply.
I know, it was getting complicated but I was experimenting and now I have a better idea how do it.
Ok with the RAM then, what about the GPU? with TrueNAS Scale I have less flexibility with the apps but at the same time it’s much simpler, any compelling reason to go one way or the other?
Why is that? Scale has KVM, k3s and Docker pre-deployed ready to run basically everything that’s on the internet.
I don"t run TrueNAS anymore. I had a TrueNAS Core VM only for storage, but I just use Proxmox for everything ZFS and a container for filesharing (SMB and Cockpit). I like to keep Proxmox hypervisor-only. Got Docker on a Rocky VM for everything container which may be expanded into Kubernetes soon, but that’s just a couple new VMs for Proxmox…
Proxmox if you want a hypervisor, TrueNAS Scale if you want everything in one box with more point&click, fancy GUI and bugs are ok.
Not sure what CPU the other server has, but Intel iGPU runs fine with Jellyfin, so no dGPU needed. I’ve also seen containers being able to share the iGPU across several pods. So there are options even without a dGPU.
I’m fine with IPMI graphics and my CPU does all the transcoding…no need for GPU.
I didn’t even look into containers within TrueNAS, I’ll check it out, it seems cleaner.
Good point about the iGPU for transcoding, I guess I was thinking more about Photoprism or Immich for face recognition, but I can always set that up on the other server that for sure will have a dGPU if the NAS doesn’t.
Thanks!
Ubuntu LXC container in Proxmox. Just samba and cockpit (for web UI) installed.
The whole “Apps” thing is just Kubernetes clustering (k3s) with Docker. Good if you want this. I’m fine with LXC and running docker stuff in a VM in Proxmox. But picking stuff from truecharts and just run it certainly has the convenience factor (if it works).
I’m now at the point where I want to go into Kubernetes, but for learning everything, I just stick to a fleet of Ubuntu/Rocky VMs and CLI