ZFS Dataset larger when made with CLI vs GUI

[Brendonb]

Hello,
I have a strange problem. I don't know if it is a feature/bug of 9.3, but I am using 9.3 so I figured it was safest to start here. Tell me if I should move this.

I have a 12 x 1TB JBOD. When I use the GUI to create a 4 x 3 raidz (4 sets of 3 disks in raidz), I end up with 10.9 available in zpool list, and 8.02TB available in zfs list. I also notice that when the GUI creates pool, it uses what I assume are gpart id's (gptid).

If I then wipe that and use zpool command to create the pool (below), I still get 10.9TB available in zpool list, but get 8.72TiB in zfs list.

zpool create -R /mnt MSA60 raidz2 multipath/disk1 multipath/disk2 multipath/disk3 multipath/disk4 multipath/disk5 multipath/disk6 multipath/disk7 multipath/disk8 multipath/disk9 multipath/disk10 multipath/disk11 multipath/disk12 log da34


Note: I have disabled the creation of swap partitions under advanced, so I assume they're not getting created, but I fear they are? I'm not great at disk geometry maths, but clearly the GUI is doing something extra that results in less space. Below is a screenshot of the difference is diskid's vs gptids (yes it is 2 different pools for example purposes).

 

[solarisguy]

I have bad news for you...

It might be difficult to find someone here who might be interested in your observation, since this is outside of FreeNAS scope. If you want to use FreeNAS and if you are wizardry enough, you do not have to use GUI to create your pools (I used to create them manually...*, since I liked them at v28). However, if your pools are outside of FreeNAS framework it is not FreeNAS anymore...

* I did everything like FreeNAS does except for the version and some unimportant from the FreeNAS point of view defaults... Swap space being zero does not enter this discussion.

 

[titan_rw]

How is this unexpected? 4 sets of 3 disks in z1 gives you the capacity of 8 disks out of 12. A single 12 disks z2 gives you the capacity of 10 disks out of 12. Why is it unexpected that the 10 disks capacity zpool is 'bigger'.

Zpool list shows the total space in the zpool including any redundancy, so it makes sense it's the same with both pools as they both have 12 drives.

As solarisguy mentions, it's best to use the gui whenever possible, so I'd stick to it.

 

[Brendonb]

Hi Titan,
It was getting late so apparently I confused the argument, but here is a screenshot denoting what I'm talking about...the first array was created with the zpool command as in the OP, and the 2nd status was created with the GUI (red arrow in pic). Both times raidz2. 700GB missing.

 

[solarisguy]

@Brendonb, please post (as code, so it is easy to see) the output of zpool history MSA60 for both cases.

 

[Brendonb]

Here is what is created by the GUI setup...

Code:

[root@azeroth] ~# zpool history MSA60
History for 'MSA60':
2014-12-02.21:44:53 zpool create -o cachefile=/data/zfs/zpool.cache -o failmode=continue -o autoexpand=on -O compression=lz4 -O aclmode=passthrough -O aclinherit=passthrough -f -m /MSA60 -o altroot=/mnt MSA60 raidz2 /dev/gptid/4196563a-7a96-11e4-b613-000743061d76 /dev/gptid/4355e324-7a96-11e4-b613-000743061d76 /dev/gptid/450fde1e-7a96-11e4-b613-000743061d76 /dev/gptid/46c961b9-7a96-11e4-b613-000743061d76 /dev/gptid/4888e084-7a96-11e4-b613-000743061d76 /dev/gptid/4a479bcc-7a96-11e4-b613-000743061d76 /dev/gptid/4c04ab3b-7a96-11e4-b613-000743061d76 /dev/gptid/4dc2eb6f-7a96-11e4-b613-000743061d76 /dev/gptid/4f799b46-7a96-11e4-b613-000743061d76 /dev/gptid/51355bca-7a96-11e4-b613-000743061d76 /dev/gptid/52eff0b1-7a96-11e4-b613-000743061d76 /dev/gptid/54a96c16-7a96-11e4-b613-000743061d76 log /dev/gptid/56694f75-7a96-11e4-b613-000743061d76
2014-12-02.21:44:58 zfs inherit mountpoint MSA60
2014-12-02.21:44:58 zpool set cachefile=/data/zfs/zpool.cache MSA60
2014-12-02.23:30:42 zfs create -o casesensitivity=sensitive MSA60/NFS-VMs

 

[Brendonb]

Here is what is produced by the CLI

Code:

[root@azeroth] ~# zpool history MSA60
History for 'MSA60':
2014-12-03.00:03:51 zpool create -R /mnt MSA60 raidz2 multipath/disk1 multipath/disk2 multipath/disk3 multipath/disk4 multipath/disk5 multipath/disk6 multipath/disk7 multipath/disk8 multipath/disk9 multipath/disk10 multipath/disk11 multipath/disk12 log da34
2014-12-03.00:03:54 zpool export MSA60
2014-12-03.00:06:07 zpool import -f -R /mnt 7569196946154629811
2014-12-03.00:06:07 zfs inherit -r mountpoint MSA60
2014-12-03.00:06:07 zpool set cachefile=/data/zfs/zpool.cache MSA60
2014-12-03.00:06:07 zfs set aclmode=passthrough MSA60
2014-12-03.00:06:12 zfs set aclinherit=passthrough MSA60

 

[Brendonb]

To be clear, I'm just trying to learn and understand what is going on here that FreeNAS does different than other ZFS systems and hopefully why. Saying "don't fiddle with CLI" seems like a brush off as there are many things that are/need to be done from CLI; why is this different?

UPDATE: this was a mistaken comment to a response I got in another of my threads on here, so apologies to direct contributors to THIS thread...I don't know how to delete an entry vs update.

 

[solarisguy]

I have no idea, but... can you retry both without using a separate log ?

 

[titan_rw]

Or try the CLI setup using partitions, the same way the GUI does it.

That's the only big thing that jumps out at me.

 

[cyberjock]

Uhh, when we tell people not to use the CLI it's more than just a brush-off. FreeNAS is an appliance. Use it as such or don't be surprised if things don't work right. As you are wanting to know the "why's" all that I'll add is that the code is on github and free to read and understand.

I hate to be the guy to brush you off, but it quickly turns into an argument that is summed up with "you should use the OS as it's designed or don't be surprised if things break". I realize you're wanting to know the "why", but that's kind of beyond the scope of the forum. We're here to support the OS in the way that it is supposed to be used. Not provide education on the intricacies of our code. Sorry if this seems harsh. We're an all-volunteer forum so most people aren't interested in spending that kind of time on questions like yours.