Zfs send should be the fastest method but it did not work on my QNAP, transfer always stopped with error after some time. So I ended up using rsync which took me like 4 weeks for 8 TB over 10G ethernet.
In theory, zfs send > rclone > rsync (with the right parameters) > rsync (defaults). I’ve never had to migrate a node over a network though, and I’m not using zfs, so can’t confirm practicality.
1 Like
what?? 
that means i am pretty much fucked, as i would need to do that twice… because i also need to move data back.
i just tried zfs send via TrueNas for a 1.5 TB dataset to a local SSD and it seems to take ages without any progress shown and it looks like its not even possible to run the task in the background (at least without scheduling it)
I agree with your ranking.
The very fastest option would probably be a block level disk cloning method (like dd or similar) for the whole disk. Then it would be a sequential process. But that probably doesn’t apply here.
For 8TB on storj I would block out two weeks and that may just be for the first pass.
I eventually used rclone it seemed a bit faster than rsync. just a bit, still slow becuase it’s millions of files.
1 Like
Maybe you should migrate to hashstore first?
4 Likes
You can run any task in the background using tools like tmux or screen:
https://linuxize.com/post/how-to-use-linux-screen/
1 Like
I also had a hard time moving last year, but eventually I stopped the node for a while and got help from hddclone…
1 Like
Basically a hardware version of dd 
2 Likes
Going back to the OP’s wish to convert from XFS to EXT4, I wonder why @Th3Van has chosen to use XFS on his newest nodes instead of EXT4 that he is using on the older nodes. I assume the nodes with the higher index are the newest nodes (http://th3van.dk/)
1 Like
I think he was testing XFS because it let you put its journal on another device (like a NVMe SSD)? I’m all for adding SSD layers in cool ways to speed up filewalker 
A good guess. So far a combination of LVM and EXT4 is working well for me. But yes, perhaps a combination of XFS hd with NVMe meta caching will perform better.