Hey everyone long time lurker here, but I’ve finally made the plunge into my first home lab. I upgraded my gaming rig and, instead of immediately selling my old parts, decided to set it up as a little home server.
Parts:
- Ryzen 9 3900x
- MSI MPG X570 Motherboard
- 32 GB DDR4 3200 RAM
- Gigabyte GTX 1080 Turbo OC
- 500 GB Samsung 980 Pro boot drive
- 250 GB SSD
- 2x 8TB Seagate IronWolf NAS HDDs
After doing some research, I decided that building out a NAS and setting up Plex/Jellyfin (recommendation?) is my primary goal right now with basic virtualization/docker support, I went with TrueNas Scale. I got a pretty good deal on the two HDDs, but they haven’t yet arrived. In the meantime, I went ahead and set up Scale to start messing around and boy do I feel like a noob.
Since I’m still waiting on the drives, I was able to set up some testing storage and get a feel for putting together Datesets and Shares - all fairly intuitive. Next, I wanted to spin up a VM, but hit a snag when Scale reported that my GPU had to be used by the host and could not be used for VMs. Unfortunately, the 3900x does not have onboard graphics, but there’s no way that Scale should need/use the 1080 to its fullest extent, will it?
After some Googling, I found that this is by design and not necessarily a bug according to iX and similarly Apps cannot use the GPU for the same reason. This made me wonder if Scale was the right solution for me after all. I’m not ready to double down and buy dedicated hardware and was really hoping to use what I had. I thought about maybe selling my 3900x and getting something with integrated graphics, but it’s a good chip and I figured it would be fun to mess around with.
What does everyone think? I did see that running Proxomox would allow me to do essentially the same thing as Scale, just with everything virtualized, but was also reading that it was a bit harder to get into.
No proxmox isn’t harder to get into but it’s different from TrueNAS.
TrueNAS is a storage system than can host apps,
ProxMox is a hypervisor which is first and foremot about compute - i.e the ability to run virtual machines and containers (LXC). It can utilise ZFS (which is the basis for TrueNAS) for VM storage but it’s not intended as network storage.
A container is a lighter way of doing a virtual machine. It shares the kernel with Proxmox (hence only being able to run Linux distros in LXC).
That said you can run TrueNAS as a VM under Proxmox or you can spin up a file server in a container and utilise the ZFS system for storage.
With Proxmox and the 1080 you could go one of a couple of ways.
Put Plex in a LXC and it will be able to access the GPU through the kernel space with Proxmox (drivers are installed at the hypervisor level). Advantage is that you still have gpu for the console.
Option 2 is to pass the GPU through to a virtual machine. This has to be a done on a 1:1 basis i.e the 1080 can only be used by one VM at a time. IF you’re interest you can go down the path of vGPU which would allow the GPU to be shared between a number of VMs.
However as part of the configuration processor you blacklist the drivers and card which means it’s not available for the console (normally won’t matter if but if you have any problems it can make trouble shooting a bit harder).
Interesting, thank you for the rundown! It sounds like I may be better off running with Proxmox? Assuming I go down that route, would I go with TrueNAS Core instead?
Why are you trying to choose between programs that are for two completely different purposes? Your question is kind of like, “Should I use Windows or GTA5”.
From my understanding, it isn’t so much Windows or GTA5 and more like car vs bicycle. If I can virtualize the NAS within Proxmox and use my 1080 for other things then my usecase is supported, having Truenas use the 1080 only for the host is non-viable for me.