Hello Fellow Unraiders, 

 

I am experiencing issues with my Unraid 7.0 server regarding a share ( What I use for my Plex files) that should use an SSD cache pool (Samsung EVO 970 ) as primary storage for new files and later transfer those files to a secondary pool (zfs_raid) using the mover. However, files are not behaving as expected.

 

Here are the issues I've encountered:

 

• Files copied to the share are written directly to the zfs_raid pool instead of the cache pool, resulting in much slower transfer speeds over my 10Gb network.
• Despite adjusting the share's settings, the expected cache-first behavior does not occur.

-----------------------------------------------------------------------------------------------------

System Configuration

 

• Unraid Version: 7.0.0
• Primary Storage (Cache): SSD (Samsung EVO 970 ) – 932GB, 11% used.
• Secondary Storage: ZFS RAID pool (zfs_raid).
• Share Settings:
  • Primary Storage: Cache
  • Secondary Storage: zfs_raid
  • Mover Action: Cache → zfs_raid
  • Minimum Free Space: Configured correctly for the cache size.

-----------------------------------------------------------------------------------------------------

 

Steps Taken So Far

 

1. Checked share settings: The Plex share is configured to use the cache pool as primary storage and the zfs_raid pool as secondary storage.
2. Verified cache directory: The directory /mnt/cache/Plex exists and contains some files, but the directory /mnt/zfs_raid/Plex also exists, and it was not fully empty initially.
3. Used the mover in test mode: Executed the mover with the following result:
4. Logs indicate the mover is working but only running in test mode (Test Mode: yes), meaning no actual file transfers occurred.
5. Disabled Docker and VMs: Ensured no other services were using the share or directories in question.
6. Manually moved files using rsync: Verified and manually moved files between /mnt/zfs_raid/Plex and /mnt/cache/Plex using:

 

rsync -avh --remove-source-files /mnt/zfs_raid/Plex/ /mnt/cache/Plex/

 

Then I tried to clean up empty directories using:

 

find /mnt/zfs_raid/Plex -type d -empty -delete

 

Although, this did not work and I cant figure out why, since there is no process using this folder now:

root@Mother:/mnt/cache# find /mnt/zfs_raid/Plex -type d -empty -delete
find: cannot delete ‘/mnt/zfs_raid/Plex’: Device or resource busy

-----------------------------------------------------------------------------------------------------

Questions

 

• Why are some files being written directly to the zfs_raid pool instead of the cache pool, even though the share is configured to use cache as primary storage?

 

• Is there an issue with my mover configuration or cache pool setup that could cause inconsistent behavior?

 

• How can I ensure the mover processes files as expected, transferring them from cache to zfs_raid only when necessary?

 

i am attaching my diagnotsics files to see if you can point me in the right direction.
 

Any help or insights would be greatly appreciated!

mother-diagnostics-20250122-2215.zip

You have no Minimum Free configured for the pool named cache or for the other pools

Plex share floor is set to over 5TB, since cache has less than that, it will write to the main pool directly.

Thanks, I did as I told and now I seem to be writing to the cache drive.

Unfortuately, the transfer speeds are really slow ( 14MB/s to max 20MB/s 😞 )  I have a 10GBe network which I have configured at MTU 4000:

 

 

 

I checked the speeds using iperf3, the results of the test seem normal to me:

 

I also checked the write speed using fio, here are the results, they seem ok:

 

Here are the results of BlackMagicDesign - Disk Speed Tests, I pointed the tester app to check my remote hard drive

 

 

 

I am attaching again my diagnostics to see if you can point me in the right direction.

 

Thanks!

 

mother-diagnostics-20250126-1946.zip

What speed do you get transferring directly to the raidz pool?

Hi @JorgeB,

Same Speed 15MB/s:

 

Thanks for your reply

That suggests a network problem, despite the good iperf results, or a client problem.

Hey @JorgeB

 

In your experience, setting up a different MTU other than 1500 is a good practice? Got this suggested by the GPT and I changed both my NICs ( Unraid and Client ) and also my Switch ( Mikrotik SPF 10gb ) to MTU 4000 to try and get the highest speed possible. 

 

Should I roll back to MTU 1500 and test speeds again?

 

Once again, thanks for your help an patience

 

If jumbo frames are supported by all the hardware involved, it can make a measurable difference with 10GbE or higher, though make sure you set the values correctly, for example for me, I use 9000 with Unraid, but in Windows have to use 9014, because you always need to add 14 for the Ethernet header.