6.12.13 Extremely slow smb writes

37 minutes ago, JorgeB said:

This is a kernel issue, but there's no fix yet for kernel 6.1, only for newer LTS kernels.

Fresh update on kernel defect

Quote

 

Christian Heusel 2024-08-27 14:44:06 UTC

Fix is queued up for 6.1 now aswell, thanks @carnil!

https://lore.kernel.org/all/2024082741-crease-mug-f658@gregkh

https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/queue-6.1/net-drop-bad-gso-csum_start-and-offset-in-virtio_net_hdr.patch

 

 

 

EDIT: In my case, the update just messed up the ethernet settings. I am running unRAID as a Proxmox VM. The VirtIO network driver was messed up. Switched to Intel e1000e and I'm getting 2.5GBE all fine. Not Lime Tech's fault!

 

Upgraded to Version 6.12.13 today. Everything is 2.5GBE. Before update, transfers were approx 2.5gbe from Windows 10 to UnRAID (SSD Cached share) and around 1.1Gbps down (always had i225 issues on the Windows Machine).

 

Upgraded to 6.12.13 and here's what I get:

Copying from Windows to UnRAID SSD Cache: says "Calculating" and it just stays there. If I wait, it times out and there's a file on the server that's the same size, but corrupted. Same if I copy to a share directly on the Array.  While it's stuck on "Calculating" I cannot close the copy window. Even browsing folders on the server slowed to a crawl.

 

I set all array drives to never sleep. No Drive actions are running. Disabled Docker and VM.

 

Pretty unusable atm.

 

 

Edited August 29 by BigJimsBoat

Same here. Extreme slow transfer speed and responsiveness when access samba from my Windows 11 PC. It works well under Mac though. After downgrading to 6.12.11, the issue went away.

In my case, all slowness issues (Windows explorer locking up during transfers) in unRiad came down to these:

 

1. My unRAID is in a VM with an HBA passed through...  I had the nic on the VM set to virtio. When I switched to intel e1000 all network issues pretty much stopped. Every VM on your stack... Check the network emulation. 

2. I tested, retested and pulled my hair out about slowness and everything locking up, until I installed Diskspeed and tested my SSD cache drives and one kept showing slow writes, very slow.  No errors. In a mirror, raid, or parity situation one slow drive screws up the entire server and will ruin your day. 

 

Next, check Windows. Update network drivers and check speeds copying to other devices 

They added a note to the known issues section of the changelog when updating:

 

Quote

 

Windows VMs should use the VirtIO-net driver

Due to a kernel regression, there is significant performance degradation for any outgoing transfer from a Windows VM using the VirtIO network driver. The workaround is to switch to the VirtIO-net driver. This should be resolved in a future release.

 

 

On 8/29/2024 at 10:13 PM, BigJimsBoat said:

EDIT: In my case, the update just messed up the ethernet settings. I am running unRAID as a Proxmox VM. The VirtIO network driver was messed up. Switched to Intel e1000e and I'm getting 2.5GBE all fine. Not Lime Tech's fault!

 

Upgraded to Version 6.12.13 today. Everything is 2.5GBE. Before update, transfers were approx 2.5gbe from Windows 10 to UnRAID (SSD Cached share) and around 1.1Gbps down (always had i225 issues on the Windows Machine).

 

Upgraded to 6.12.13 and here's what I get:

Copying from Windows to UnRAID SSD Cache: says "Calculating" and it just stays there. If I wait, it times out and there's a file on the server that's the same size, but corrupted. Same if I copy to a share directly on the Array.  While it's stuck on "Calculating" I cannot close the copy window. Even browsing folders on the server slowed to a crawl.

 

I set all array drives to never sleep. No Drive actions are running. Disabled Docker and VM.

 

Pretty unusable atm.

 

 

Yesterday I had to upgrade one of my Unraid servers running under Proxmox and hit the same issue, but I then got hit with new issues (call traces etc) with E1000.
Ended up using vmxnet3 and its been stable overnight.

The patch for this is out in the wild now, hopefully I'll just upgrade to unraid 7 (with a newer kernel) and be done with the issue.

Edited Sunday at 11:07 PM by tjb_altf4