I rebuilt my 2x1 TB NVME cache pool from BTRFS to ZFS after version 6.12 came out. Now (after several weeks) i noticed, that mover action was incredibly slow. Even if running mover without any filters (or mover tuning plugin) a move of about 100 GB of media data needs more than 2 hours. Which is way too slow in my system. In BTRFS this was about 15 minutes or so - just maxed out by the write speed of the array at about 60 - 70 MB/s.  Before i switch back to BTRFS: Has anyone an idea where to look or optimize. Because the r/w performance to cache is excellent, when using ZFS cache pool.

All ZFS tests here showed up bad results ("slow as a dog"), if in pools or in array drives.

I've reverted to XFS and all is fine again now...

 

Thx. I'll switch back.

48 minutes ago, MAM59 said:

if in pools or in array drives.

There are known issues with zfs in the array, but not with pools.

6 hours ago, JorgeB said:

the array, but not with pools.

I've seen them here, believe it or not. ZFS is a total flop with unraid.

17 hours ago, MAM59 said:

I've seen them here, believe it or not.

Likely some other issue, I'm using like 12 zfs pools without any issues, and pool to pool copy was even a little faster that TrueNAS, not by much, but certainly not slower.

I didn't remember if I still had this but a few months ago I tested with both Unraid and TrueNAS, copy from one zfs pool to another, using the exact same hardware and settings, copy was done using Windows explorer but they were done locally using Samba's server side copy, no network involved:

 

 

 

I would say, it depends on the hardware used. In some environments, it works, in others its the worst thing you can do to your server.

So everybody should be prepared to see a huge degragation of speed and think of a rollback.

(I have never said, that it cannot work ANYWHERE, I just say "here" its really a mess)

 

ZFS hat nur 8 GB cache bei mir...  bei 64 GB ram