Missing Space: need to deleted an unusepool and use it to extend another

[philjans]

Hi,
I have been looking for doc and video about this "simple" task but since it is a production environment, I want to besure.
I am missing space in one of my pool (RipCluster) but I have another one (ShareDaniel) that can be deleted and the freed space can go the the 1st one.

I see the option "Deleted Dataset".

My first question is why is it call pool and dataset at the same time? If ShareDaniel is a pool, I want to delete "a pool"...

Then, once the delete is done and I have 2.91tb of free unassigned space, I don't see the option to "extend pool":

Thanks for the information!
Really appreciated

 

[JaimieV]

ShareDaniel is a dataset, not a pool. From the pictures above, the only pool shown is Tank. There is no inconsistency, ShareDaniel is always described as a dataset in the pictures above.

One or more disks make a vdev,
one or more vdevs make a pool,
a pool contains one or more datasets,
a dataset contains a filesystem or block data, and perhaps more datasets.

Deleting a dataset doesn't result in spare disks you can bring back; it deletes all the content of the dataset, so you will get the freed space of the released 24.82GB of data currently in ShareDaniel that your first two pictures show. The freed space will go back to the pool, available to all datasets. It's not much though, RipCluster already has considerably more free space than that.

RipCluster must have a quota set, as its free space does not equal the free space on the pool (Tank says 7.89TB available in the pool). Just go into "Edit Options" for RipCluster and increase the quota to 'make space' (actually you are "raising the ceiling") on RipCluster. The other datasets also seem to have quotas set.

Since this is a production environment, go find someone who knows how it works before you do anything like deletion. If you've inherited this and it's now yours to look after, learn how the backups work before touching it. Go look in Tasks/Replication and Tasks/rsync, and Tasks/Snapshots too. See what's in there, read the manual about those pages.

 

[philjans]

Hi,
Thanks for the great and clear answer. Yes I am looking at going through the trainings but our employees hit the quota and fixing this was the priority.

So if I change on RipCluster:
"Quota for this dataset and all children" from 2.93 to 3.93: they will instantly get 1tb more space without loosing data?
Do I have to also change "Reserved space for this dataset and all children" to 3.93tb?

Thank you very much.
Phil

 

[JaimieV]

Yes, RipCluster will get another Tb instantly. Be aware that some software might inappropriately cache how much free space it has and not expect it to increase, I don’t know if whatever’s using RipCluster does but if you don’t get the expected result, try restarting the software in the appropriate way.

The “reserved” setting will avoid other use of the Tank pool eating into the RipCluster space, so probably yes, if RipCluster takes priority.

 

[philjans]

Thanks for the info.
RipCluster is used only by Proxmox so we will see if it accepts the added space.

 

[philjans]

... I don't know if it is related but as soon as a added 1tb, I got this message:

But this dataset is not limited:

And no quotas:

 

[philjans]

Warning gone. Proxmox saw the space right away. All is good. Thank you for your great help.

 

[JaimieV]

No probs. ZFS supports an effectively unlimited number of iNodes (expands on demand) so I've no idea what the error might have been caused by.