ZFS overhead question

[John_n4s]

Hi there everyone,

I am new to the forums, but have been using FreeNAS for more than 5 years now. I have a specific question regarding storage in my raidz1 (4 x 8TB) encrypted pool.

When I am in the WebUI it shows the following stats:
Total RAW space:
14.7 used + 14.3 avail = 29 TiB

Total data space (minus parity):
10.7 used + 9.8 avail = 20,5 TiB

According to this calculator: https://jsfiddle.net/Biduleohm/paq5u7z5/1/embedded/result/
I should have about 21,83 TiB of usable data space.
Thus I seem to have 21,83 TiB - 20,5 TiB of overhead.

Looking a little further I find this calculator as well: http://wintelguy.com/zfs-calc.pl
Which has a 'slop space allocation' button that makes the calculation come close to my actual values. (see attached screenshot)

Is it normal to have 1,33 TiB of overhead/slop space?
Which calculator is correct?
Am i missing something?
Thank you for your time.

 

[Dice]

Is it normal

yes.

 

[John_n4s]

yes.

Thanks for the reply Dice!
Do you happen to know what this 1,33 TiB is?
I know it is listed as 'slop space" on the last URL but what that means is still unclear to me.

 

[Johnnie Black]

I noticed the same in my new 8 x 3TB RAIDZ2, about 1.5TiB less than I expected, apparently it's due to ZFS padding with 4k sector disks, some configurations have much less overhead, e.g., 6 disk RAIDZ2

 

[mrsieb]

Well same story here....

12x8TB-RaidZ2,

During creation estimated space ~72TB , already bitty that in fact a 8TB Disk have effective 7,2TB on storage
(caused by the case that harddisk manufactures calculate with 1000 bytes not with 1024)
but at the end my Pool end ab at 64TB available space additional 8TB are vaporized....

is there something to what i can optimize ?

BR

 

[Dice]

are vaporized

I feel you. All of you.

Think of the lost TBs as "invested" in data security :)
Once the initial cost anxiety has settled, you'll progress to the next stage.
At some point you'll only quickly check by biduleohms calculator to get an idea of how much storage your next vdev would grant your pool. Then you get what you get. That is what matters. Not any other comparisons to days on EXT4 or NTFS.
The sooner the better one <uninstalls> any conception of how much storage you get based on glancing the sticker on the drive, the better.

welcome onboard.

 

[Stux]

Well same story here....

12x8TB-RaidZ2,

During creation estimated space ~72TB , already bitty that in fact a 8TB Disk have effective 7,2TB on storage
(caused by the case that harddisk manufactures calculate with 1000 bytes not with 1024)
but at the end my Pool end ab at 64TB available space additional 8TB are vaporized....

is there something to what i can optimize ?

BR

That doesn't sound right to me :-/

You should have about 72TiB available. Which is simply 10x7.2TiB, or basically 80TB.

FWIW, I've gotten over the whole TB/TiB HD conspiracy.

You should plan your upgrade at 80% full and upgrade by 90% full. You can use more than 90% but the array will slow down drastically

 

[rs225]

FWIW, I've gotten over the whole TB/TiB HD conspiracy.

The interesting part is storage(and network or clockrate) has never been sold in base 2, but 'everyone' knows they switched to base 10 and invented a perpetual rip-off machine.

 

[Stux]

Only thing that was ever base2 was RAM, and perhaps Core Memory

In fact, I think FreeNAS UI should start switching over to TB instead of TiB, then the problem goes away where someone buys a pair of 8 TB drives and wonders why they only get 7.2TiB.

For bonus points, make it a configurable option to display either TB or TiB.

 

[mrsieb]

That doesn't sound right to me :-/
You should have about 72TiB available. Which is simply 10x7.2TiB, or basically 80TB.

Well... what to say...

I also wonder if it would be better to make 2 mirrors with 6Disk RaidZ1 instead off 1 mirror with 12Disk RaidZ2...
BR

 

[Dice]

I also wonder if it would be better to make 2 mirrors with 6Disk RaidZ1 instead off 1 mirror with 12Disk RaidZ2...

For the sake of data integrity, using large drives, this is a no-go.

 

[Stux]

Well... what to say...

I also wonder if it would be better to make 2 mirrors with 6Disk RaidZ1 instead off 1 mirror with 12Disk RaidZ2...
BR

The problem is that the 7.2TiB "parity and padding" has risen to 20TB with double parity. I would've only expected it to raise to circa 16TiB

And in fact, in an 8-way, you get 16.755TiB.

FWIW, I wouldn't recommend a 12-way Z2 vdev.

For 12 drives, you might as well go to RaidZ3, and still have 61.3TiB available. Alternatively, 2 6 way RaidZ2 vdevs, for 56TiB and double the IOP performance and much better redundancy.

I didn't realise the parity/padding increased so rapidly once you got beyond 10-way.

9-way seems to be fairly optimal, with 15.5TiB for Parity/Padding.

All assuming that the calculator you used is accurate :)

PS: are your vdevs encrypted?

 

[Bidule0hm]

To answer the original question: please read this thread https://forums.freenas.org/index.php?threads/misaligned-pools-and-lost-space.40288/ it should make things clear ;)

 

[John_n4s]

To answer the original question: please read this thread https://forums.freenas.org/index.php?threads/misaligned-pools-and-lost-space.40288/ it should make things clear ;)

That was exactly what I was looking for. Very nice description :)

 

[mrsieb]

The problem is that the 7.2TiB "parity and padding" has risen to 20TB with double parity.

For 12 drives, you might as well go to RaidZ3, and still have 61.3TiB available.

PS: are your vdevs encrypted?

Well I am dealing around the first time with "big" disks beyond 3TB so % counts even more ;-)
Infact i upgrade my nas from formal 16 Disk a 2TB to 12 a 8TB and still not sure what is the "best" setup
for my case 90% big media files.

yes i switched encryption on allready durring creating the mirror means befor creating a dataset..

To answer the original question: please read this thread https://forums.freenas.org/index.php?threads/misaligned-pools-and-lost-space.40288/ it should make things clear ;)

nice thx!