It's like asking if hauling a fifth wheel with your truck is any safer now than it was years ago. You are physically doing about the same thing, so the overall risks are quite similar. It is still a big, heavy, potentially tippy, hard to back up arrangement that will lose badly if you do any number of things.
Some improvements have been made. These are core technological improvements, similar to how electric brakes or rearview cameras are things that are common today that weren't common years ago. They ease some of the pain points like mountain hauling or parking the beast, but don't change the overall physics of the situation or the general danger involved.
One of the biggest problems a decade ago was that PCIe passthru was a nascent technology, and the implementations on many systems were very poor and unreliable. This problem has definitely gotten vastly better; I wouldn't EXPECT a broken implementation to be a problem if you buy a server-grade platform that was designed in the last five years. I have chosen those words carefully, though. The server-grade boards are the ones most likely to get all the i's dotted and t's crossed correctly. I have definitely seen Sandy and Ivy systems with PCIe passthru weirdness, and maybe the occasional Haswell or Broadwell. I have no recollection of having run across people with newer stuff than that with issues on name brand commodity systems (i.e. Supermicro, HP, Dell). These companies are able to dump sufficient engineering effort into their systems to make obscure features like PCIe passthru work correctly.
Likewise, picking VMware as your hypervisor platform is the best path to success. While some people have had success with Proxmox, the PCIe passthru support there is still listed as experimental and was only introduced maybe five years ago. It has been problematic for some people. VMware, on the other hand, rakes in big bucks from their enterprise licensing and has put a lot of polish on their support for PCIe passthru.
So those are definite improvements, but you are still not guaranteed success.
Regardless of the date you see on the post, or the fact that it was posted when the product was still named "FreeNAS", the information I posted at
[---- 2018/02/27: This is still as relevant as ever. As PCIe-Passthru has matured, fewer problems are reported. I've updated some specific things known to be problematic ----] [---- 2014/12/24: Note, there is another post discussing how to deploy a small FreeNAS VM instance for basic file...
www.truenas.com
still represents an accurate summary of the state of the art in safely virtualizing TrueNAS. Ignore bullet points at your own peril. If you truly know why you can ignore them safely, good for you, go right ahead. My goal is simply to make sure you know the nature of the potential minefield you're walking into.
There is no such thing as "just as safe". There are inherently additional risks with virtualization. Even if you follow my guidance, for example, it is entirely possible for your automated deployment of ESXi to misfire and overwrite one of your TrueNAS data disks. You are blending together two incredibly complicated software systems, one of which is generally intended to be hosting somewhat simpler virtual machines, the other of which was designed to be running on bare metal. You are pushing the boundaries and it would be foolish to proceed without awareness of the sharp edges present.
