I have on vm with a disk size of 6TiB. Everytime I need to reboot the host, or change the vm's configuration, the bitmap is dirty causing the next backup task to do a full disk read. This is very slow. It takes 15 hours aprox. to do the whole disk. Not much is being synced to the remote PBS however, because not much data has actually changed.

Is there a way to speed up the backup task? Are my disks just too slow? The VM disk is a "raw" vm disk on a ZFS partition, containing 2 mirrors of 8TB hard drives.

Normaly a reboot don't create a dirty bitmap, but a stop and start does, so don't stop the instance... And don't backup on differents target neither..

Normaly a reboot don't create a dirty bitmap, but a stop and start does, so don't stop the instance... And don't backup on differents target neither..

I know that, that's why I said: restart of the host, or changes to the vm config (it needs to be shutdown/stopped for that some of the changes to take effect).
But it's "normal" that a backup of a 6TiB drive would take 15 hours or longers? What exactly does it do? Read the block and compare them to what's already on the PBS? So I'm guessing my drives are just not that fast?

Hard disks could be your issue as long as you are not limited by 1G network anyway. Transfer of 6TB over 1G network would take about 13 hours, that is pretty close to those 15 hours. But without details I am just guessing.

Hard disks could be your issue as long as you are not limited by 1G network anyway. Transfer of 6TB over 1G network would take about 13 hours, that is pretty close to those 15 hours. But without details I am just guessing.

Well, the PBS is a VM itself, and it sits on same same PVE host. However, the PBS storage is "remote" as in I've added it as a PBS store inside PVE. So I'm guessing it's just going over the bridge of PVE to the PBS vm.
But the reading of the vm disk should not involve any networking I'm guessing? What would be the best way to stress test the ZFS volume?