NPIV (vHBA) on proxmox

[aarcane]

Is there a way to enable NPIV or vHBA on Proxmox, or is this on the roadmap? I need this for an expansion I have planned, and SCSI device passthrough is unlikely to be sufficient.

The plan is to build a new SAN host with two distinct vSAN arrays, and connect them both to the Fibre Channel network. They will be exporting ZFS volumes as LUNs for other physical and virtual systems around the LAN, using a mix of iSCSI and FC. (FC booting on Windows is amazing. ZFS iSCSI for Proxmox looks pretty cool). The two vSAN arrays will have significantly different pool profiles and performance requirements, and for that reason, I want to segregate them into two different VMs running on the same dedicated PVE host. The only shared resource I haven't found a documented solution for is the FC problem. Specifically, PCIE Passthrough (Non SR-IOV/NPIV based passthrough) is NOT a solution, as the best suitable host I can find has 3 PCI-E slots, one HBA for each VM, and a shared 4 port FC adapter, with 10GBE on mezzanine that supports SR-IOV.

 

[LnxBil]

I'm also wondering if this is going to be a thing or not. I tried it years ago, but I could not get it to work. In the end, it was then not fully implemented in QEM/KVM. PCIe passthrough was the only thing worked for me, and it worked fine. I also exported ZFS datasets directly via FC including trim-support.

best suitable host I can find has 3 PCI-E slots,

That's bad. Normally this is not the case for server hardware, e.g. the box i use currently has 10x PCIe ports, all at least x4 mechanical,