Understanding replicated snapshot sizes

[f00b4r00]

Hi,

I have two FreeNAS servers (see signature), and I'm replicating the snapshots from one to the other. I know and understand that due to the differing parity scheme between the two (Z2 vs Z1) the actual used/referenced sizes may differ somewhat due to parity data.

What really bugs me though is this:

Code:

Source FreeNAS:

poolain/Video@auto-20200503.0015-2m    2.34G    -    4.38T    -
poolain/Video@auto-20200524.0015-2m    410K    -    4.41T    -
poolain/Video@auto-20200614.0015-2m    367K    -    4.42T    -
poolain/Video@auto-20200621.0015-2m    341K    -    4.43T    -
poolain/Video@auto-20200628.0015-2m    0    -    4.43T    -

NAME                                                         AVAIL   USED  USEDSNAP  USEDDS  USEDREFRESERV  USEDCHILD
poolain/Video                                                3.42T  4.43T     2.40G   4.43T              0          0


Replication target:

poolette/Replik/PeroPero/Video@auto-20200503.0015-2m    51.9G    -    4.47T    -
poolette/Replik/PeroPero/Video@auto-20200524.0015-2m    302K    -    4.45T    -
poolette/Replik/PeroPero/Video@auto-20200614.0015-2m    262K    -    4.39T    -
poolette/Replik/PeroPero/Video@auto-20200621.0015-2m    232K    -    4.39T    -
poolette/Replik/PeroPero/Video@auto-20200628.0015-2m    0    -    4.38T    -

NAME                                                           AVAIL   USED  USEDSNAP  USEDDS  USEDREFRESERV  USEDCHILD
poolette/Replik/PeroPero/Video                                 3.51T  4.52T      139G   4.38T              0          0

In other words: while the data is _growing_ on the source server (I'm adding data to the Video dataset), the referenced data is _decreasing_ on the target, and that really boggles my (simple) mind. FWIW the target dataset is read-only, a pure recipient for the periodic replication task snapshots.

Can someone shed some light onto this? Did I just create antimatter compression? ;P
Thanks