Dear all

 

After i tested UNRAID 7 on my testserver without any problem I updated my main server as well from 6.12.6 to UNRAID 7.0.1 last weekend. 

The update went smoothly but now I discover much slower connection to my server. 

 

iperf tells me no problem, but if I send something from or to my server by SMB on my Windows11 PC I got "just" around 350MB/s. Before the update I had around 800Mb/s without any problem for both sides (send & receive).

 

The same issue I have if i transfer from cache to cache drive with krusader. Before the update I had about 300MB/s and now only around 150 MB/s. 

I have not changed anything about the HW. Just updated UNRAID. 

 

Another problem I faced after the update, the cache function was not properly tranfered. It was writing directly to the array and not to the cache. After I just set the primary storage again to cache it was working well again. - solved

 

Does anyone have an idea wat i can do to get the old speed back? Or can I go back to the old version of UNRAID?

 

Thank you very much in advance. attached the diagnostic file

kronos-diagnostics-20250505-2224.zip

Do the speeds get better if you downgrade to 6.15?

I have not checked it if I can downgrade again. 

if this is possible, I can check it. But it definetly was better before the upgrade

So finally i tested the downgrade. 

 

Everything is on the same speed level as before the update to UNRAID 7.0. 

 

- Transfer speed in Krusader 2x faster 

- Transfer speed by SMB to the server 2x faster

 

I can not understand where this is coming from. Any Ideas?

I see you are using a Mellanox NIC, not aware of any issues with them and 7.x, I use them myself, run a single stream iperf test in both directions with 6.12 and 7.x to rule out any network issues.

I checked this already, the iperf3 has similiar results.

I think something about the speed for the cache disks is wrong. Cause I have also a lower speed if I transfer something from disk to disk within krusader

That's very strange, but I'm afraid I don't have more ideas, haven't seen any similar issue reported.