Since a while one of my pve hosts runs a zfs dataset with native encryption and did not have any issues so far.
I've not implemented automatic loading of the key for the encrypted dataset so if the host is booted I will have to supply it myself.
Now I was testing snapshots on this dataset and it seems to work fine.
However when I was monitoring if data is counted in the zfs snapshot stats, I saw when doing a
Snapshot on unmounted encrypted dataset:
Off course before starting the vm, first mount the dataset supplying the key, then starting the vm.
USED and REFER stats stay the same, even when changes are made inside the vm.
Snapshot on mounted encrypted dataset:
Start the VM.
USED and REFER stats are increasing.
I wondered if this difference could cause any issues.
Snapshots were done in the web gui, but that probably doesn't matter.
pveversion: pve-manager/7.4-3/9002ab8a (running kernel: 6.2.9-1-pve)
I've not implemented automatic loading of the key for the encrypted dataset so if the host is booted I will have to supply it myself.
Now I was testing snapshots on this dataset and it seems to work fine.
However when I was monitoring if data is counted in the zfs snapshot stats, I saw when doing a
zfs list -t snapshot, a difference appears when snapshotting a vm using this storage if its mounted or unmounted.Snapshot on unmounted encrypted dataset:
Off course before starting the vm, first mount the dataset supplying the key, then starting the vm.
USED and REFER stats stay the same, even when changes are made inside the vm.
Snapshot on mounted encrypted dataset:
Start the VM.
USED and REFER stats are increasing.
I wondered if this difference could cause any issues.
Snapshots were done in the web gui, but that probably doesn't matter.
pveversion: pve-manager/7.4-3/9002ab8a (running kernel: 6.2.9-1-pve)