all space gone in ZFS

[asw2012]

tl;dr
If you destroy a snapshot, anything that was deleted or overwritten in the past will not be recoverable. If you destroy all your snapshots, you will be stuck with your data as it currently stands - no recovering old files.

I mean that, I've personally had situations where I thought my data was complete but Resilio Sync actually got rid of some files and when I went back to look for them (obviously not there) I resorted to the snapshots only to find that I didn't have snapshots going back far enough to recover the data. I then had to resort to my external backup to hope that the data was there.

that is what I thought.

Well, I've destroyed all of the snapshots and 6.1TB magically appeared after I rebooted the system. The snapshots were the culprit again it seems..

 

[diedrichg]

So does that mean you have all the free space you are expecting or is there more missing?

 

[asw2012]

So does that mean you have all the free space you are expecting or is there more missing?

it means I have all of the free space I am expecting.

 

[asw2012]

Thanks everyone for the help. Dang snapshots bit me again.

 

[fracai]

"zfs list -o space" would have been of great help here as it lists how much space is taken up by snapshots per dataset

 

[Spearfoot]

"zfs list -o space" would have been of great help here as it lists how much space is taken up by snapshots per dataset

The OP posted the output of zfs list -o space in his original post, but it didn't show the space used for snapshots in the /mnt/RAIDz2/Share directory, which was not a dataset, but was configured as a share. Seems like a bit of a 'glitch' in the way ZFS presents snapshot usage...

 

[SweetAndLow]

The OP posted the output of zfs list -o space in his original post, but it didn't show the space used for snapshots in the /mnt/RAIDz2/Share directory, which was not a dataset, but was configured as a share. Seems like a bit of a 'glitch' in the way ZFS presents snapshot usage...

The snapshots are in the Raidz2 dataset. Seems correct

Sent from my Nexus 5X using Tapatalk

 

[Spearfoot]

The snapshots are in the Raidz2 dataset. Seems correct

Sent from my Nexus 5X using Tapatalk

The command only reported 645MB as USEDSNAP for RAIDz2 with 15T for USEDDS, but the OP ended up finding ~6TB of space used by snapshots in RAIDz2/Share, which didn't show up anywhere at all in the zfs list -o space output -- because it wasn't a dataset. The 6TB more-or-less 'fell through the cracks'.

 

[SweetAndLow]

The command only reported 645MB as USEDSNAP for RAIDz2 with 15T for USEDDS, but the OP ended up finding ~6TB of space used by snapshots in RAIDz2/Share, which didn't show up anywhere at all in the zfs list -o space output -- because it wasn't a dataset. The 6TB more-or-less 'fell through the cracks'.

Not possible. If there are snapshots they are on the dataset. Let me look back through things and maybe i can explain.

 

[Spearfoot]

Not possible. If there are snapshots they are on the dataset. Let me look back through things and maybe i can explain.

Yes, the snapshots were on the pool; they just weren't reported anywhere by the zfs list -o space command, other than implicitly in the USEDDS value (15TB).

 

[fracai]

Yes, the snapshots were on the pool; they just weren't reported anywhere by the zfs list -o space command, other than implicitly in the USEDDS value (15TB).

Yeah, that's weird. I would have thought they should have shown up in the snapshot column.
I know reporting space can be weird if the data is referenced in multiple places, but if the space was cleared by deleting all the snapshots, that should have been tallied in the snapshot column as well.

 

[diedrichg]

Did we ever learn which version FreeNAS was being used? Could there have been a build with a bug in the zfs list -o space string for that build?