Losing large chunks of disk

[LostFate]

This may have been reported on (I'll be the first to admit that I'm lazy). This also may not be a bug (Though in all my research, I haven't come across anything about it).

So I just detached a disk, marked it as new, and then made a new pool out of it. The new pool lists the total available disk space as 928.0GiB. Immediately after, it lists a default dataset created with the pool... 899.0GiB available. I checked to verify that the obligatory swap partition wasn't set to an absurd amount (The default: 2GiB).

Is FreeNAS setting a buffer so I don't overfill my drive? Something else entirely? Bug? I mean, in the grand scheme, 29GiB is not the end of the world (however this is but one of my pools. I imagine that I am taking a somewhat more substantial hit).

Just curious about what might be going on.

 

[dlavigne]

This: https://forums.freenas.org/index.php?threads/dumb-volume-dataset-question.25765/#post-164017.

 

[LostFate]

Motherboard: ASUS Crosshair 4 Extreme (This is a testing platform, before I get blasted about the extremely inferior hardware.)
CPU: AMD Phenom 2 x4 965
RAM: 8GB DDR3

FreeNAS Version: FreeNAS-9.3-STABLE-201412200530
Onboard Disk Controller Settings: AHCI
Swap: 2 GB per pool

Pool 1:

Name: Disk_Images
Disk 1: Western Digital - WD4000F9YZ-09N20L0
​

Pool 2:

Name: Media
Disk Layout: Striped
Disk 1: Seagate - ST32000641AS
Disk 2: Seagate - ST32000641AS
​

Pool 3:

Name: Jail_Data
Disk 1: Western Digital - WD740GD-00FLC0
​

Pool 4:

Name: Misc_Storage
Disk 1: Hitachi Deskstar 7k1000
​

Specifically, I am asking about the discrepancy between Misc_Storage and its subsequent Dataset. They were generated by the Volume Manager and have not been touched in anyway since creation. Why is the Dataset, which I imagine is created to fill the entire drive, reporting 29 less GiB than the pool? Looking at the others shows that they all lose a pretty sizable amount of storage in this way (Though the others are loaded with data so who knows...).

 

[dlavigne]

See https://forums.freenas.org/index.php?threads/dumb-volume-dataset-question.25765/#post-164017.

 

[LostFate]

Thank you for your remarkably lazy answer to my admittedly lazy question. I've gone through that thread several times now and can, with a rather sound level of certainty, tell you that it doesn't in anyway even begin to touch on the subject which I've brought up. My question isn't "HEY!... I HAVE A TB DRIVE, WHY AM I ONLY GETTING 928GB!!!!!1!!!"... It's "Hey... I have a TB drive. I should be getting 928 GiB. I'm only getting 899. What gives?". For all I know, ZFS as a file system imposes a massive amount of overhead on disk space. It could be using that for legitimate purposes and not be a bug. I'm simply asking, where has the 29GiB gone?

 

[danb35]

Keep in mind, when mentioning "massive overhead", that you're dealing with about 3% of your pool's capacity. Despite your certainty that my post (that Dru linked to above) is irrelevant to your question, I think it may still be the answer, though less obviously so since you aren't using any sort of RAID on this pool.

Be warned that I'm veering into speculation--I don't know anything about the internals of FreeNAS. However, this speculation fits what I've seen on my own system, and what I've seen described by other people with similar questions. It's my understanding that the volumes page lists the "volume" (i.e., pool) capacity as the total raw capacity of the device(s) comprising the pool, and the dataset (every pool is also a dataset) capacity as the net available storage capacity of that dataset. Most of the time, the pool is using some flavor of RAID, and the capacity loss to redundancy dwarfs any formatting/filesystem overhead. But even without RAID, the filesystem takes some overhead to do its thing. I've seen a Youtube video of Dru giving a talk about ZFS (perhaps this one: ), in which she mentions that ZFS takes 1/64 of the capacity for its own purposes, like metadata backups, checksums, etc. That would account for 1.6%, or just over half, of your "missing" space. I honestly don't know where the remainder would be, but I suspect it's something similar.

Now, as I said, this is speculation, as I've never seen this documented explicitly. If this is how it works (or even if it isn't), it'd be good to have it documented. But I think it's a reasonable guess as to the answer to your question.