dataset / pool size shrinking

[jkmurphy1]

I recently rebuilt my Freenas system. Previously I was running 9.1.x (in a virtual environment --the reason for the rebuild). While rebuilding I updated to 9.2.8. i didn't go to 9.3.x because I was having difficulties getting AD settings correct and I was on a time constraint to have the rebuild completed during off peak hours (couldn't get AD users and groups to import).

The array is small currently, only two 2TB drives, but will have more disks added once I'm confident I do not need to revert back to the previous system. The new system has been in use for ~ 3 days. The size of the pool started at ~1.8TB. It is down to 1.2TB as of right now.

When running 'zfs list', the pool appears as I would expect it to (used + available = size). However, in the gui and on a windows machine, the shares from the storage pool appear to have a size equal to the available space, not the allotted size. This can be seen in the attached image.


The only additional space consuming feature that is set up on the machine is zfs snapshots, which is creating 2 snapshots daily.


My main concern is running out of space, or the appearance of running out of space to my end users may be more accurate. Is there a way to correct this? If this isn't corrected, what kind of implications will this have once my actual used space exceeds the incorrect size that is being reported (or the reported available space drops below 0)?

 

[danb35]

Did you set up snapshots? That would explain the used space.

 

[jkmurphy1]

Yes, snapshots are being taken 2 times daily of one of the datasets. I was under the impression that snapshots only captured changes. I really want to get this figured out. Since posting this the size of the primary share has decreased by 3GB.

 

[Bidule0hm]

It's because of HDD's rats: "HDD rats are long-tailed very small rodents hidden in the drives and who eat small pieces of the platters."

 

[jkmurphy1]

How very helpful

 

[Bidule0hm]

Just kidding a bit, sorry :P

Actually on my pool I have 0.2 TiB missing since the pool was empty (was 12.2 TiB when empty, now 2.0 TiB used and 10.0 TiB free) but I think it's some overhead and not related to your problem because you lose space even if you don't touch anything, it's right ?

 

[jkmurphy1]

Not entirely. I have ~ 30 users pinging and saving to this throughout the day. But the 3 GB change that has occurred today is not normal usage and not the problem I am having. The daily snapshots are normally under 100MB if that gives an idea of new file storage.

The concern is I originally started out with a pool of 1.8TB. It is now down to 1.2TB after ~4 days up and running. I am wondering if 9.2.8 is still affected by the bug reported here https://bugs.freenas.org/issues/4419 or if my issue is the same as what is reported in this bug.

 

[Bidule0hm]

Ah, I see, you're on 9.2.

On the 9.2 version the total size of a dataset (including the root dataset) was calculated with [total space] - [sum of the used space on all the others datasets] so for example:

- You have a 5 TB empty pool
- You create 2 datasets: /pool/abc and /pool/def
- You add 0.1 TB of data in /abc
- You add 0.5 TB of data in /def

You'll have these dataset sizes:

/pool/ 4.4 TB (5 - (0.1 + 0.5))
/pool/abc 4.5 TB (5 - 0.5)
/pool/def 4.9 TB (5 - 0.1)