As someone still in monolithic home server tech debt I've been meaning to migrate too. The part I always wonder about is storage - said server is a NAS, serving files over SMB and media over Plex. From what I hear some people mount and export their data array [1] directly on Proxmox and only virtualize things above the storage layer, like Plex, while others PCI-passthrough an HBA [2] to a NAS VM. I suppose one advantage of doing the former is that you might be able to directly bind mount it into an LXC container rather than using loopback SMB/NFS/9p, right?
I also hear some people just go with TrueNAS or Unraid as their bare-metal base and use it for both storage and as a hypervisor, which makes sense. I might have to try the Linux version of TrueNAS now that it's had some time to mature
I also rely on my Intel processor's built-in Quicksync for HW transcoding. Is there any trouble getting that to run through a VM, except presumably sacrificing the local Proxmox TTY?
[1] As in the one with the files on, not VHDs. I default to keeping those separate in the first place, but it's also necessary since I can't afford to put all data on SSDs nor tolerate OS roots on HDDs. Otherwise a single master ZFS array with datasets for files and zvols or just NFS for VM roots would probably be ideal
[2] Which I hear is considered more reliable than individual SATA passthrough, but has the caveat of being more coarse-grained. I.e. you wouldn't be able to have any non-array disks attached to it due to the VM having exclusive control of all its ports
I also hear some people just go with TrueNAS or Unraid as their bare-metal base and use it for both storage and as a hypervisor, which makes sense. I might have to try the Linux version of TrueNAS now that it's had some time to mature
I also rely on my Intel processor's built-in Quicksync for HW transcoding. Is there any trouble getting that to run through a VM, except presumably sacrificing the local Proxmox TTY?
[1] As in the one with the files on, not VHDs. I default to keeping those separate in the first place, but it's also necessary since I can't afford to put all data on SSDs nor tolerate OS roots on HDDs. Otherwise a single master ZFS array with datasets for files and zvols or just NFS for VM roots would probably be ideal
[2] Which I hear is considered more reliable than individual SATA passthrough, but has the caveat of being more coarse-grained. I.e. you wouldn't be able to have any non-array disks attached to it due to the VM having exclusive control of all its ports