[SOLVED] Enable sparse on existing ZFS storage

[limone]

Hi,

I noticed my manually created zfs pool is not sparse (thin).

Is there an easy way to activate it afterwards?
I could easily add "sparse 1" to my storage config, but would that work, and most important, not destroy the data?

 

[fabian]

that would only affect newly create volumes/datasets. but you can easily set the reservation for existing ones manually as well - see 'man zfs'

 

[limone]

I've set it to sparse 1 now, and migrate the disks from that datastore to another and then back, so it's a new disk, right?
(zfs set reservation=0 NVMe/vm-540-disk-1 doesn't change anything)

 

[limone]

Ok looks like migrating to another storage and back does not do the trick

before:

root@pve-lab:~# zfs get all NVMe/vm-580-disk-0 | grep used
NVMe/vm-580-disk-0  used                  10.3G                  -
NVMe/vm-580-disk-0  usedbysnapshots       0B                     -
NVMe/vm-580-disk-0  usedbydataset         7.66G                  -
NVMe/vm-580-disk-0  usedbychildren        0B                     -
NVMe/vm-580-disk-0  usedbyrefreservation  2.66G                  -
NVMe/vm-580-disk-0  logicalused           7.62G                  -

after:

root@pve-lab:~# zfs get all NVMe/vm-580-disk-0 | grep used
NVMe/vm-580-disk-0  used                  9.96G                  -
NVMe/vm-580-disk-0  usedbysnapshots       0B                     -
NVMe/vm-580-disk-0  usedbydataset         9.96G                  -
NVMe/vm-580-disk-0  usedbychildren        0B                     -
NVMe/vm-580-disk-0  usedbyrefreservation  0B                     -
NVMe/vm-580-disk-0  logicalused           9.93G                  -

 

[fabian]

you probably need to trim the volume afterwards to regain unused space. just setting the refreservation without moving the disk at all would have done the trick as well.

 

[limone]

Unfortunately that didn't work either

root@pve-lab:~# zfs get all NVMe/vm-901-disk-0 | grep used   
NVMe/vm-901-disk-0  used                  82.5G                  -
NVMe/vm-901-disk-0  usedbysnapshots       0B                     -
NVMe/vm-901-disk-0  usedbydataset         12.9G                  -
NVMe/vm-901-disk-0  usedbychildren        0B                     -
NVMe/vm-901-disk-0  usedbyrefreservation  69.6G                  -
NVMe/vm-901-disk-0  logicalused           12.8G                  -

after `fstrim /` inside the VM, it's still the same

 

[limone]

I've set it to sparse 1 now, and migrate the disks from that datastore to another and then back, so it's a new disk, right?
(zfs set reservation=0 NVMe/vm-540-disk-1 doesn't change anything)

The correct command is "zfs set refreservation=0G NVMe/vm-901-disk-0"

my bad
*facepalm*