RAID-Z1 vs a degraded RAID-Z2

[gtrrmsd]

I know you can't currently convert a RAIDZ1 into a RAIDZ2. I would rather run RAIDZ2. I have 3 x 3TB disks. So I can make a 2+1 RAIDZ1, or I can make a degraded 2+2 RAIDZ2 and add another disk later when I can afford it.

What are the implications, is there any difference, of running a degraded 4 disk RAIDZ2 with only 3 disks vs running a optimal 3 disk RAIDZ1?

 

[Yatti420]

you need to run the proper amount of required disks.. 3 disks for a raidz1 and 4 disks for raidz2.. aka you can't create a raidz2 with 3 disks..

Please review http://doc.freenas.org/index.php/Volumes#ZFS_Volume_Manager

 

[gtrrmsd]

I wont be trying to create a 3 disk RAIDZ2, I'll be creating a degraded 4 disk RAIDZ2 with 3 disks or creating an optimal RAIDZ1 with 3 disks, but I'm not sure if there is any implications between the two.

From the surface it looks like they are the same, both 3 disk arrays, with 6TB capacity, and able to survive one disk failure, but I'm thinking there may be some other differences? (speed is not a large issue for me)

http://forums.freenas.org/index.php...dz-3-of-4-drives-i-e-to-allow-migration.7748/
(I can also borrow another 3TB disk to create a 4 disk RAIDZ2 and then pull and return it.)

 

[cyberjock]

You don't want to operate with a degraded array. FreeNAS makes it a priority to email you and whatnot when things aren't right. If you've deliberately broken them to begin with then you have no choice but to expect the daily emails (and then realize they tell you the pool is degraded and realize you don't know if it's expected or not) or to actually disable them. In either case you're playing with fire.

My recommendation is to either stick with a 4 disk RAIDZ2 or a 3 disk RAIDZ1 (RAIDZ2 is obviously better).

 

[gtrrmsd]

I don't mind daily emails, or keeping track of if the RAIDZ2 is having actual problems. Those two things I would happily trade for the ability to be using an optimal RAIDZ2 down the track when I get another disk, vs having to copy off the data, destroy, remake, and re copy the data to move from a 3 disk RAIDZ1 to a 4 disk RAIDZ2 later.

Aside from it being a kind of messy way, are there any technical differences between the degraded 4 disk RAIDZ2 vs the optimal 3 disk RAIDZ1?

 

[cyberjock]

I don't know. It's not something I've seriously considered. It's like buying a car with only 3 tires because "someday i'll want a 4th". Its just not the kind of thing I'd endorse and considering that's not how its intended I think you are just asking for this to backfire since it's not tried and true.

I'm really confused on why you're not just waiting until you can have a real full RAIDZ2. People go with ZFS because of it's amazing data integrity checks. Why would you want that, but then turn around and do something like deliberately operate with a degraded array. It just doesn't compute. Either you care about data integrity and your decisions reflect that or you don't care and your decisions reflect that. I think that you are trying to claim that integrity is important, but then simultaneously disregarding it in other ways. I think you need to consider going full-on and doing it right or choosing to use something else. But that's just my 2 cents.

 

[gtrrmsd]

I would say it's more like buying a car without a spare wheel, with the plan of someday putting one in the trunk. :D If I search 'convert RAIDZ1 to RAIDZ2' there are certainly many many hits.

 

[cyberjock]

Yeah, and there's a small tutorial here on how to create zpools with sparse files. It hasn't ended well to the people that tried it, so it makes me think this is just waiting to pwn you and your data. Add to that the fact that if this goes bad you can't run any of those standard recovery tools to get your data back and you'd be really screwed. I think you should stick with what you currently use until you are ready to make a proper jump to ZFS.

 

[gtrrmsd]

What are the commonly used recovery tools relating to ZFS? ZFS disaster recovery is definitely something I will be reading up on over the weekend.

What can go bad though vs just having a 3 disk RAIDZ1? Both can loose one disk, and not a second. If something can go bad worse when using a degraded RAIDZ2 than in a 3 disk RAIDZ1 then there must be differences in how ZFS implements RIADZ1 and RAIDZ2?

I wont have these three 3TB disks until next week anyway, I have boxes of older disks here, so I might set up a test server here with some ~500GB disks or something and test it out a bit.

 

[cyberjock]

What are the commonly used recovery tools relating to ZFS? ZFS disaster recovery is definitely something I will be reading up on over the weekend.

There are none. You restore from backup. (This is mentioned in my noobie guide)

What can go bad though vs just having a 3 disk RAIDZ1? Both can loose one disk, and not a second. If something can go bad worse when using a degraded RAIDZ2 than in a 3 disk RAIDZ1 then there must be differences in how ZFS implements RIADZ1 and RAIDZ2?

I wont have these three 3TB disks until next week anyway, I have boxes of older disks here, so I might set up a test server here with some ~500GB disks or something and test it out a bit.

You don't understand my concern. What you are wanting to do is far outside the realm of the "proper" or "expected" way to manage a server. If your data is important you should be a little more than hesitant to do anything of this sort.

You're welcome to test all you want. That still doesn't change my recommendation. We've seen plenty of people test weird things and it seems to be okay, but once they do it in production it goes badly. Unfortunately many of these people didn't have backups and the pool ending up corrupted also meant they lost all of their data.

Yes, this means that for some people there's been lots of tears and heartache. If you want to take the risk, it's yours to take. I know I'd never consider doing what you are considering with *my* data.

 

[gtrrmsd]

Nope I don't quite understand your concern :D

I'm really confused on why you're not just waiting until you can have a real full RAIDZ2. People go with ZFS because of it's amazing data integrity checks. Why would you want that, but then turn around and do something like deliberately operate with a degraded array. It just doesn't compute.

You are implying that the 3 disk RAIDZ1 and the degraded 4 disk RAIDZ2 with only 3 disks are different, and that the degraded RAIDZ2 is worse / more dangerous. But apart from it reporting that it is degraded as you mentioned, I don't understand why they are different? Both have parity and can sustain 1 failed disk, and I assume both have all the same data integrity metedata checksums etc etc.
(One other difference I can see is that there will be overhead creating the parity, but that is not an issue for me.)

 

[enemy85]

Nope I don't quite understand your concern :D



You are implying that the 3 disk RAIDZ1 and the degraded 4 disk RAIDZ2 with only 3 disks are different, and that the degraded RAIDZ2 is worse / more dangerous. But apart from it reporting that it is degraded as you mentioned, I don't understand why they are different? Both have parity and can sustain 1 failed disk, and I assume both have all the same data integrity metedata checksums etc etc.
(One other difference I can see is that there will be overhead creating the parity, but that is not an issue for me.)

I think that CJ wants to warn you that creating a degraded raidz2 from CLI is not suggested because it will end up in a mess while creating a raidz2 with a borrowed disk pulled immediately out after creation (making the system work in a degraded state for normal use) is not raccomended either because u won't understand if the system as other problems apart from that.