Confused with the filesystem zfs vx xfs and arrays vs pools

[Ronald Q]

Hi im totally new to unraid. I have my server up and running checking parity now with a trial key. But so many doubts.

 

I have 15 HDDS (all 2tb hitachi sata)  and 2x120gb ssd. Old hardware but want to make a Complete home NAS solution and some docker containers.

 

1) whats the difference between the main array and the pools? If i use my 15 hhds in the main array i can only run raidz2? what about raidz3 in the main array?

2) What is the point to have an Array in the first place? why not to create single/multiple pools instead?

3) Can i expand later with more disks a raidz2 array/pool?

4) if i replace the parity drives with bigger ones, then i can replace one of the 2 tb array drives with a bigger one?

5) best file system for my 2xssd (sata3) in cache pool?

6) Main differences between ZFS and xfs file systems? future expandability? performance?

7) Will my dockers run better in the ssd (mirrored)?

 

[JorgeB]

Array uses parity or dual parity for redundancy with any filesystem, raidz is only for zfs pools, each has advantages and disadvantages, most are here:

 

https://forums.unraid.net/topic/131857-soon™️-612-series/?do=findComment&comment=1243850

 

[primeval_god]

22 hours ago, Ronald Q said:

1) whats the difference between the main array and the pools? If i use my 15 hhds in the main array i can only run raidz2? what about raidz3 in the main array?

2) What is the point to have an Array in the first place? why not to create single/multiple pools instead?

This can be a point of confusion for many new users. The term Array will hopefully soon be a legacy term (the term not the functionality).  The term "Array" refers to the currently singular, currently required pool that uses the proprietary unRAID redundancy scheme. Here is a more in depth post I made to try to explain the concept a while back 

 

[itimpi]

@primeval_god Perhaps it is time to agree on new nomenclature and start using it?    Perhaps we need to start talking about something like an "unraid pool" or perhaps a "hybrid pool" or even "unraid hybrid pool" (to avoid confusion with 'ZFS hybrid' pools)?   Has anyone seen a suggestion from Limetech on what their preference is likely to be?

[JorgeB]

I believe Tom mentioned somewhere that it should be called unRAID, you can then add pool or array to that, the OS is still Unraid.

[primeval_god]

My understanding was that at some point the "array" terminology would be replaced with something like "unRAID pool". So the options for disk pool types would be xfs (single disk but still referred to as a pool), BTRFS (1 or more disks using BTRFS raid levels), ZFS (1 or more disks using ZFS raid levels), unRAID (1 or more disks, mixed files systems, using Limetech's proprietary RAID like redundancy). My hope would be that Limetech updates the terminology coincident with whatever release removes the requirement for having at least one disk in the unRAID pool.

 

p.s. What is a "ZFS hybrid" pool, i am not familiar with the term, or zfs in general.

Edited June 3 by primeval_god

[itimpi]

5 minutes ago, primeval_god said:

p.s. What is a "ZFS hybrid" pool, i am not familiar with the term, or zfs in general

Nor am I - it just came as a result of a Google search.

[JorgeB]

13 minutes ago, primeval_god said:

What is a "ZFS hybrid" pool

That usually means a zfs pool with subspool(s), like a special (metadata) vdev for example.

[Ronald Q]

Thank you all for the reply. According to what i understand now after a lot of research/trial and error, this work like this:

 

1) The "main array" ONLY can use proprietary unraid redundancy and easy expansion scheme. So, You can have here 1 or 2 parity drives max (regardless of the filesystem you use). also this main array is mandatory, at least until actual release 6.12.10. So you need to have at least 1 disk on here, you use it or not. 

 

Advantages of the unraid main array:

• easy expansion adding more disks without need for parity sync or array rebuild due to expansion.
• you can use different size disk without loosing any space
• if you loose a disk you only loose access and performance on the files stored in that specific disk until it is replaced and new disk filled again with the parity data. Other files in the "unraid main array" will not be at risk in this process, and performance to serve those files would not be degraded.

Disadvantages on the unraid main array:

• speed is not great with spinning drives because you are not stripping the data along all the disks in the array, instead you write files to a single disk and the parity disks.
• you cannot have 3 parity drives (max 2)

2) The so called pool is an array (using the legacy array term form the hardware RAID era) or collection of disks. In those pools we can configure disks to make arrays (mirror, 1 stripping 0, raid 10 type, single disk parity, dual disk parity or triple disk parity) depending on the file system you use in those pools. Something similar to the vdevs used in trueNAS software.

 

advantages of the unraid pools:

• you can create super fast raidZ arrays or vdevs using mechanical drives. For example with 24 disks you can create 4 groups of 6 disks. Each group/vdev can be single parity raidz1, dual parity raid z2, or triple parity raidz3.
• you can create pools with different filesystems that suit best the specific requirements of the files, dockers or vms you will run/store on those pools.

disadvantages of the unraid pool:

• question: performance will be degraded in case of a failed drive? until you replace failed disk and rebuild? im not entirely sure about this. specially if you use raidz2 or raidz3 models.

 

3) definitely there is something pending to do with the terminology in the documentation. Is very confusing at first for new users. Please don't get me wrong i like this softwarea lot. But im used to the traditional hardware raid terminology.

 

4) what i miss actually:

• ability to assign a single/subpool drive like ssd/nvme for cache to a specific pool or the main unraid array for exclusive use. If your network don't bottleneck you can use a raid10 of nvmes to cache a spinning drives pool. So depending on your hardware you can achieve 24.000 or more mbits/s of more in write speeds to that pool. This can open uraid to a new level of enterprise customers. The only disadvantages i could find on youtube for unraid are lack of performance for large or enterprise applications.
• being able to not use the main unraid proprietary array if i don't need it.
• global l2arc cache

Please correct me if im wrong on anythig.

Thanks

 

 

 

 

[primeval_god]

2 minutes ago, Ronald Q said:

• if you loose a disk you only loose access and performance on the files stored in that specific disk until it is replaced and new disk filled again with the parity data. Other files in the "unraid main array" will not be at risk in this process, and performance to serve those files would not be degraded.

This is not correct. If a device in the "unRAID array" fails and you have enough redundancy to rebuild it then there will be no loss of access as the disk can be emulated using parity and the data from the other disks in the array. There is a performance penalty in accessing data from an "emulated" disk, as well as a penalty when accessing the array in general while a rebuild is happening (when you are replacing a disk). This is pretty the same way that a traditional RAID system works. 

 

The real advantage of the unRAID array type is what happens if you lose more drives than you have redundancy. In the case of a traditional RAID system if you lose more devices than your redundancy you lose the entire array. For an unRAID array since each disk (except for parity) has a stand alone file system you only lose the data on the disks that have failed. Any functional data drives can be removed and mounted as a normal linux file system. 

[Ronald Q]

thanks a lot for your reply. Understood