Recovering usable space?

Operating Systems & Open Source
,
1

Hi all,

I want to clarify if I move data out of a dataset (as I have a separate vdev that I can use), once the replications outlive their retention period, does ZFS “reclaim” the available free space?

Right now, my primary system is between 89-90% usage, and I’m getting warning reports.

Cheers,
Mike
(also, Merry post-Christmas!)

2

Hi Mike! Merry Christmas!

If your replication snapshot expires and is removed, any actual data stored on the pool is freed. This amount can be seen in the “refer” column on zfs list.

That said, refer for a snapshot only considers old revisions of files or blocks that were updated on the main dataset. Basically, the snapshot tells zfs to not free the original block when copy-on-write does its thing. This means that if the dataset is largely unchanged, the snapshot might not actually be taking up any space. In order for me to give you a concrete answer, it would help to see the output of zfs list -t all.

2 Likes
3

Happy new year too mate!

Yes, it looks like a lot of the snapshots are just “tags” as my use these days is mainly retention and just sync to the backup host. The vdev shows used space as 46.1T.

pool1/new_dataset                                             28.1T  3.82T  28.1T  /mnt/pool1/new_dataset
pool1/[email protected]                       66.4M      -  27.9T  -
pool1/[email protected]                       64.8M      -  27.9T  -
pool1/[email protected]                       64.8M      -  27.9T  -
pool1/[email protected]                       64.8M      -  27.9T  -
pool1/[email protected]                       64.9M      -  27.9T  -
pool1/[email protected]                       64.9M      -  27.9T  -
pool1/[email protected]                       65.2M      -  28.0T  -
pool1/[email protected]                       65.2M      -  28.0T  -
pool1/[email protected]                       65.3M      -  28.1T  -
pool1/[email protected]                       65.3M      -  28.1T  -
pool1/[email protected]                       65.3M      -  28.1T  -
pool1/[email protected]                       65.3M      -  28.1T  -
pool1/[email protected]                       65.4M      -  28.1T  -
pool1/[email protected]                       65.4M      -  28.1T  -

One option is to move this largest dataset to a new pool, but my issue there is the extra pool I have is just a 8-drive vdev as raidz2 (without a sync copy) like this primary vdev.

  • Moving it will clear up the available space once the 2w window passes
  • Cons: if there’s a 3-drive failure on the new pool, that dataset is lost.
4

Based on that table, it doesn’t look like you’re going to recover any space from removing those tags.

How big are the individual member disks? As long as they’re sufficiently new and not too massive, i think z2 is plenty. That’s what i have on my NAS, after all.

Apologies for the slow response. New year is always wild