Can you create a ZVol with a drive missing and then add it later?

[JoeS]

Hi,

Can I create a RAIDZ2 volume with 3 3TB disks and one missing disk (by creating a loopback device to null or something) and then use this volume in a degraded state? Could I then just add the missing disk at a later date? Is there any downside to doing this?

Thanks,

Joe

 

[cyberjock]

You can... but its incredibly stupid. We've had some users that did it and it failed miserably for them. The reason for the problem is not understood. They simply saw the remaining devices in the pool become corrupted resulting on the pool being lost.

Just don't do it. The manual doesn't say anything about doing it, so don't try to cheat the system. You'll be sorry. ;)

 

[JoeS]

Ok, I'll have to save up some cash for an extra drive. In that case if you have a 4 drive RAIDZ2 ZVOL and one drive fails then don't you encounter the same problem?

Joe

 

[JoeS]

All of this data is backed up so did the corruption occur after the missing disk was replaced or during the time that the volume was degraded? If its the latter then I don't mind taking the risk.

Joe

 

[cyberjock]

No. What happened was the 3 disk RAIDZ2(1 disk missing) worked fine. But then one day they rebooted the server or did something that shouldn't matter and on reboot their pool was corrupted.

Like I said.. don't do it and just don't even think about doing it. What you are really doing is creating a redundant pool that seems to corrupt itself randomly.. ergo there is no real redunancy.

Just don't do it...mmmkay? :D

 

[JoeS]

Hmm, surely its the same scenario as if I was to just buy 4 disks, create the pool, then have one disk fail and then replace that disk? I understand that if you run the pool in a degraded state then it may be a bit unstable whilst it is degraded, I can accept that as a risk since everything is backed up. However if your saying that the above scenario can result in the pool randomly failing at some point in the future then that calls the whole point of ZFS in to question. Surely ZFS is designed to allow you to recover from a hard drive failure?

Do you have any links to cases where this has happened or any syslogd logs or anything? Does anyone else have an opinion on this?

Thanks,

Joe

 

[cyberjock]

Hmm, surely its the same scenario as if I was to just buy 4 disks, create the pool, then have one disk fail and then replace that disk?

That's exactly what I would think, but it doesn't work out that way.

I understand that if you run the pool in a degraded state then it may be a bit unstable whilst it is degraded, I can accept that as a risk since everything is backed up. However if your saying that the above scenario can result in the pool randomly failing at some point in the future then that calls the whole point of ZFS in to question. Surely ZFS is designed to allow you to recover from a hard drive failure?

Yes, but ZFS makes assumptions like... "The server admin won't try to stupidly undercut me" and "The server admin won't stupidly virtualize me" and "I will have total hardware access". We already bbq people for virtualizing, we bbq people for using hardware RAID thereby denying total hardware access. So naturally why would we expect it to work? It's just like ECC RAM. Either you do it right or you do it wrong. ZFS doesn't care what choice you do, but your data will if you lose it.

Do you have any links to cases where this has happened or any syslogd logs or anything? Does anyone else have an opinion on this?

Thanks,

Joe

Honestly, its not worth my time to even go searching for it. It's your data and if you don't want to believe me then you shouldn't bother reading the manual either since I've made some critically important corrections in the manual too. We don't try to identify problems where people do stupid things to lose their data and we don't try to recover data from situations that are clearly a lost cause(virtualizing, null drives, etc.). Those threads aren't going to tell you much except they lost their data. It's not worth our time to "try to fix stupid".

If you think its a great idea go for it. But it's stupid from my position and from my experience. If you want to save $130 on a hard drive right now in exchange for possibly losing your data go for it. I won't stop you. But you'll get no sympathy if you decide to start doing things that aren't in the manual. And I can tell you for 100% certainty that it does NOT recommend you use null devices.

 

[JoeS]

Ok well I'm going to go and buy all of my hardware now except the drives. I'll spend a month messing about with various configurations etc and then buy all 4 drives when I next get paid.

You mentioned that virtualization is not recommended, I understand why but this is something that I need to do because of money and space constraints. This is for a home environment and as I said everything is backed up. One question that I have is that I notice that a lot of people on here prefer Intel (I didn't go the Intel route because since I am virtualizing everything I need a CPU with as many cores as possible.) What I was wondering is if VT-d has any kind of advantage over AMD-vi? I would assume that as far as FreeNAS is concerned the passed through HBA looks and behaves the same way regardless?

Joe