New to Scale, Size of my Pool is not as expected (does not change with recordsize)

[Kira]

Hello I am new to TrueNas an ZFS.

I build a machine with two pools
The first pool is 10x20TB as RaidZ2.
Size shown when creating the Pool as estimate is 145.51TiB (Which ist 8*20TB).
But the automatically created Dataset is only 138.42TiB. I then set the Recordsize of the Dataset from 128k to 1M in the hopes the overhead will be lower (read something like that on the internet with calculations for it). But still no changes.
According to OpenZFS Capacity Calculator the overhead should shrink from 4,8 to 0,48%. Which is a lot smaller and usable capacity should be 144,819TiB.
It is also of no change if I use encryption or not.

The second pool is a Pool of 2x3.84TB SSD (3,48TiB) as Mirror which come out with only 3.38TiB usable capacity (which was expected by the sheet).

Is this a bug in the dashboard or did i do something wrong? I cannot set recordsize when creating the pool but later only change it on the dataset. Which does nothing for capacity somehow.

[ABain]

Hoping this resource will help, lots of background

The problem with RAIDZ

This resource was originally created by user: Jamberry on the TrueNAS Community Forums Archive. Please DM this account or comment in this thread to claim it. The problem with RAIDZ or why you probably won’t get the storage efficiency you think you will get.​ Work in progress, probably contains errors! Better formatted github version. As a ZFS rookie, I struggled a fair bit to find out what settings I should use for my Proxmox hypervisor. Hopefully, this could help other ZFS rookies. This te…

[Kira]

Hello,
thanks for the input. This resource is new although I found various on ZFS Sizing and Overhead.
TlDr: “So is TrueNas estimate of space on Dataset always based on 128k Recordsize even when 1M is set?”

Mainly I don’t look at absolute storage efficiency because this is (for me) sometimes misleading when comparing between Z2 and Z3 so I only looked at ZFS Overhead.

I used several resources (including the linked which is also very good) and came from the sheet and my calculations (according to sources that describe it like yours).
That all boils down to the 10disk RaidZ2 being 4,8% of a loss with Sectorsize 4k (ashift=12), Recordsize 128k. Absolutely normal what TrueNAS is showing (138.424 TiB).
But I expected TrueNAS to show 144.819 TiB when changing the Recordsize to 1M for the Dataset. Changing ashift to 9 is no solution since 4k is the real sector size of the drives.

All in all I think there are 2 possibilities:

1. TrueNAS does only show an estimate based on default 128k Recordsize and when selecting 1M the Data takes up less space, so usable space is more than with 128k. (btw I know the size is an estimate and you cannot compare the values to a filesystem like ext4) => this is what I think is actually happening
2. TrueNAS does not take the 1M value and uses 128k and the estimate is based on what TrueNAS does. (this is more unlikely)

So I have two options minimizing overhead. Change to 1M and bear with the displayed size or go for a RaidZ3 with 128k which should only have an overhead of 0.189% and usable size of 145.244 TiB

So I am to decide if I buy another Drive to go RaidZ3 (which I was wrapping my head around the whole time) or just choose 1M and live with whats shown.

The Size of the 2 SSD Mirror contrary is not really “healable” so I have to live with it. I bought these because 2x7,68TB and up where beyond my financial scope (along with buying 10x20TB and telling my wife about it). But this poses no real problem.