Hyper-V support as standard in new images

[Barry_J]

Hi guys,

I've had a pretty rough time after upgrading to 9.2.1.7 and then 9.2.1.8 getting my Time Machine backups to work. I was formerly using 9.2.1.3 as a VM on Hyper-V and didn't run into hardly any issues at all until I ran out of space on the disk I was using. I decided to change the disk and give FreeNAS an upgrade as well (as it had been 6 months without an update) and found that I couldn't get my 10.9.5 Mavericks install to back up anymore.

Anyway, after trawling the internet, I found an interesting PowerPoint from some Microsoft Devs at from this year's TechEd and found that you have actually already made a Hyper-V supported VM whilst they have begun their quest to support FreeBSD. It might only be 9.2.1 (available here: http://download.freenas.org/9.2.1/HYPERV/RELEASE/) but after having enough with the 9.2.1.8 issues and AFP just not playing nice anymore, I decided to try it out. Not only is all the Hyper-V hardware supported out of the box but it is also blazing quick compared with having to use the legacy network adapter option (when I say super quick, I was getting between 3-7MB/s transfer from the Mac before and now I'm getting 35 - 45MB/s using the standard Hyper-V Network Adapter). To the inevitable trolls: this is an eco single disk only and that's OK for me since it's in a VM so please hang fire on the sarcastic "mine's better than yours" comments,

So back on point; I just wondered if it would be too much to ask if Hyper-V support can now be added as standard to the images since it's a brilliant performance boost and allows people like me who already have a server to not have to buy more Apple only kit for a single purpose.

Thanks for your help!

Barry

 

[cyberjock]

Actually, 9.2.1 is the only one with Hyper-V support. Not privvy to *all* of the reasons why no others were released, but I know part of it is because we don't want to give the impression that virtualizing FreeNAS is a good idea. It's not. We've seen time and time again that it ends badly. I know another reason is that Microsoft intially started working with us, then did a "we'll get back to you in a month" and that was like 8 months ago.

In fact, in my previous post from less than a minute ago I just had to explain to someone that supporting and endorsing a given configuration is two totally different things. They aren't the same, but the risks are so high with virtualizing (and the consequences so sudden and so severe) that it just didn't make sense.

But don't let me end the discussion. I have no doubt other people are going to show up and say something like "Virtualizing FreeNAS is for losers" or something because virtualizing is serious and nothing to joke about. There's a reason why our forum rules even go so far as to say "don't expect responses on these topics if questioned" and lists virtualizing FreeNAS as one of them. ;)

It's crazy dangerous, it's crazy stupid without backups, and it's something that we can neither stop nor ever endorse with a clean conscience.

 

[jkh]

The truth is somewhat more prosaic. I did the HyperV/9.2.1 release as an experiment which wasn't repeated simply because the required code was on a branch and it was too problematic to keep it up-to-date (anyone else wishing to build their own hyperv images need simply check out the feature/hyperv branch, sync it up to date as desired, and build their own FreeNAS).

I consider virtualization in general somewhat inevitable, regardless of whatever cyberjock may say or feel to the contrary, even though FreeNAS is currently designed to work on native hardware without virtualization interposed. Why? Because it's just too convenient, virtualization technology is steadily getting better, and there will probably come a day when even standard consumer/enterprise PCs ship with a hypervisor and "only fools" actually try to talk to the hardware directly (which may, in fact, no longer even support the notion of native OSes due to TPM / secure boot technology and other BIOS interaction issues which simply make running under virtualization the only way to go). When that day comes, FreeNAS will be just one of many software appliances.

In any case, when FreeNAS 10 is released, HyperV support will come along for the ride along with the VMWare support that is already there. In point of fact, FreeNAS has already shipped with the vmware tools necessary to make it interoperate properly with vmware for a number of years now, so claiming that HyperV isn't supported because "virtualization is bad" would be a rather inconsistent and hypocritical argument to make - that's not the reason at all. We simply haven't gotten around to HyperV in an official capacity yet and have no plans to do so until FreeNAS 10.

 

[cyberjock]

Interesting.. not what I had heard, but okay. I'll take it from the horse's mouth FTW.

 

[mjws00]

@jkh That was way too prosaic. Heh. ;) But also nice to see a forward looking opinion. Thanks.

 

[Barry_J]

@cyberjock; thanks for your contribution. I have no experience really with the FreeNAS system but I do have experience with enterprise virtualisation on many different hypervisors (even got a few badges to prove it :D) but I do find it strange that you say so many people have had it fail in virtualisation. Do you know why this is just out of interest? My only reason for it failing previously was probably because of my lack of experience with the system and the fact that I tried to change the disks it was using for the Time Machine storage; rather than adding to the ZFS volume. I know that for years, doing a number of things on virtualisation was considered 'bad'; starting with database servers, and progressing on to directory servers because of issues with delay writes, memory ballooning implementations, inappropriate use of snapshots and other strange behaviour in these applications that just left people feeling 'it ain't as good as good 'ole single OS on a physical system'. That being said; I don't really entertain this as a valid argument anymore and I'm really glad to have had @jkh's contribution too because it's great to see that FreeNAS are considering that this could be one of a number of appliances running on a nice piece of kit that shares its compute power and has passthrough access to some physical disks. For me to put FreeNAS on a physical system by itself just seems a bit wasteful whilst I've got a Xeon based server I've built that has more power than I've ever been able to consume perfectly able to emulate other things. The question was around support in future releases which I've gratefully got my answer to from @jk. However your experience around virtualisation is still interesting to me nevertheless and I will just continue to cautiously use this as the Time Machine backup system I originally intended and stick with keeping the important data managed by Windows Server for now until I'm in a position to invest in more hardware.

Great discussion though and it's been really encouraging to see some serious knowledge titans on this just in case I do need some help in the future!

 

[cyberjock]

I think you guys are far too optimistic. The problem isn't that we don't want FreeNAS to work in a virtualized way, it's that it doesn't work, and the problem isn't entirely clear. The one problem we're fairly sure of is that the problem is on the virtualization side of the house.

My arguement against virtualizing isn't because of wants or needs, or even just because I want to be a jerk. It's because objective quality evidence has come to the conclusion that it's just not a good idea. If that changes next year, then great. The problem is that it's not likely to change in the next year. FreeNAS can be recoded all day long, but if there are gaps or flaws in the virtualization layer, it's likely we'll never be able to fully understand them let alone work through them (assuming they can be worked through).

As for the exact problem, we don't know. We've had people that reboot their ESXi server for maintenance and their pool is gone. No clue why as they didn't do anything that most people would call "bad". We just know, from experience on the forums, that many many users have created virtualized environments for FreeNAS and everything worked great. Then they'd do something very innocuous like shut the server down for a weekend and when it boots back up their pool is gone.

If we actually knew what the problem was we might have a workaround. The problem is nobody has been able to provide a consistent way to reproduce the problem, and you might do whatever kills your pool 20 times without a problem. But that 21st time went badly for whatever reason.

 

[mjws00]

The random nature of it grinds my gears more than it should. I'm totally down for "it does not work, because (x)." That I can deal with, workaround, decide it's not worth it. But I don't even know how to attack the problem, besides run it, and try and break it. Unfortunately the risk makes the product a non viable solution for clients. Which is a crying shame as a version that is stable under esxi is a killer app, imho. Simple DC, file services protected with zfs, time machine, backup server... Man I can put that everywhere I am p2v'ing or replacing an old windows server.

I am pretty optimistic that the bugs will get worked out, even if we never know them. Realistically a future hardware platform, updated hypervisor, or just updated FreenNAS or BSD version, has the potential to just "disappear" whatever unknown problem occurs. But I am firmly in the virtualization will be ubiquitous camp. There are just so many advantages to a thin layer, whether for security, scalability, or compatibility. It gets better every day, it's built in to our chips at a hardware level, and the money being thrown at it as a solution is mind-blowing.

It is a tough one to reconcile. I trust FreeBSD under esxi (naively?), it doesn't randomly dump zfs pools. I've run all the other native zfs platforms without fear or failure or being evil for virtualizing. I'm almost to the point of trusting ZoL, but meh. FreeNAS feels like it is so close I can taste it, and I like it way more.

I want to believe. ;)

 

[ravensorb]

It would be nice to see at least the core network and disk drivers added -- this would at least allow for us to use it in small test/dev environments.

 

[jkh]

Aren't those already merged into FreeBSD10.2? If everything I understand about Microsoft's recent work with FreeBSD is true, HyperV support should just come along for the ride.