Hey there unRAIDers.

 

 

I am currently working on my third unRAID server, so not completely new to the system.

However this setup is a bit different as it is designed to act as my main PC, with VMs running in order to support multiple graphics cards and up to 7 monitors at a time.

 

It is bleeding edge hardware, which seems to be causing some issues. I am hoping someone here can help me figure out the issue, or if all I can do is wait for next release with updated kernel support...

 

The issue at hand is that the integrated MB NIC seems to randomly cut out, leaving my with no network access to the web interface, and also no network connection on my VMs.

I have been trying everything I can think of, and also bought two USB NIC based on RTL8153, in order to attempt bypassing the built-in NIC. I'll get back to those.

The issue seems to occur randomly, and can be after either several hours to a day or so.

When it happens everything looks fine on the settings and such, except on the interface portion of the dash, where the connection speed is simply missing.

My priority is trying to get the integrated NIC working, with the USBs as a secondary solution, more on that further down. This screenshot is with the integrated NIC when the issue is occuring.

 

I have also confirmed that the NIC itself is without fault, as there is no issues when booting directly into W11, bypassing unraid.

 

 

Using the USB NIC:

I have installed the Realtek RTL8153 driver package, which should support the USB NICs. When booting the system I have internet very briefly, but then it cuts out again immediately with no recourse.

Currently I am doing a passthrough to my main VM of the USB controller, in order to ensure a working connection on the Windows VM, as this is also my primary work PC. And I can confirm that the NIC work without a hitch on here.

 

I am at a loss here. Attaching diagnostics file as captured moments ago, with the issue being active.

 

And yes, I have tried replacing cables etc.

 

Unraid version: 6.11.5

 

System hardware:
MB: Asus X670E-E GAMING WIFI

CPU: AMD Ryzen 9 7950X3D

RAM: 4x16GB DDR5 G.Skill RAM

GPU: PNY GeForce RTX 4090
GPU2: Asus GT730

 

theone-diagnostics-20230313-1356.zip

Edited March 13, 20233 yr by TalkingWaterfall

first of all: throw away those USB Nics, they are a pain in the a@$. For no apparent reason, they block when data hits them fast enough and they never release the block. Its a driver issue. The funny thing is, they work if you pass them on to a windows VM, as you have already noticed 🙂

 

The onboard NIC is a delicate 2,5Gbs type. Yeah delicate, because 2,5 does not really exist, its a marketing thing. Physically its a "10GBs with 3 Pauses". Some switches recocnize these cards as 10Gbs and overrun the line. A working "flow control" on both ends can help, so be sure you have enabled it in the switch (UNRAID automatically turns it on).

And, despite of what the marketing department will tell you, USE PROPER CABLES!

 

You may check it out by limiting the card to 1G only. If it runs without problems over some time, you have found the culprit.

 

12 minutes ago, MAM59 said:

first of all: throw away those USB Nics, they are a pain in the a@$. For no apparent reason, they block when data hits them fast enough and they never release the block. Its a driver issue. The funny thing is, they work if you pass them on to a windows VM, as you have already noticed 🙂

 

The onboard NIC is a delicate 2,5Gbs type. Yeah delicate, because 2,5 does not really exist, its a marketing thing. Physically its a "10GBs with 3 Pauses". Some switches recocnize these cards as 10Gbs and overrun the line. A working "flow control" on both ends can help, so be sure you have enabled it in the switch (UNRAID automatically turns it on).

And, despite of what the marketing department will tell you, USE PROPER CABLES!

 

You may check it out by limiting the card to 1G only. If it runs without problems over some time, you have found the culprit.

 

Thanks for the input so far, very enlightening. Do you have some input as to how I can accomplish this limit? I have an Asus GT-AX11000 Router and a "dumb" switch splitting the ports. Been looking around in unRAID and my router interface with no luck.

The cables are very high quality, cat8.

 

I'll be scrapping the idea of using the USB NIC as backup, and just pass them through to the VMs

28 minutes ago, TalkingWaterfall said:

The cables are very high quality, cat8.

yeah, also a marketing gag. The cables maybe ok, but get suspicious if they sell it declared as "raw cable".

This would mean, the cables are CAT8, but the plugs are not!

CAT8 plugs are still new in the shops, almost all patch cables that you still can buy are NOT capable of running high speed without problems.

 

Yeah I know, there is no "do not autospeedselect" option in the UNRAID GUI. I usually limit the switche's port to 1G then. But with a "dumb" switch, this is no option too, sorry.

It should be possible with the "ethtool" command, but I am not sure, which syntax is needed.

 

 

Edited March 13, 20233 yr by MAM59

6 minutes ago, MAM59 said:

yeah, also a marketing gag. The cables maybe ok, but get suspicious if they sell it declared as "raw cable".

This would mean, the cables are CAT8, but the plugs are not!

CAT8 plugs are still new in the shops, almost all patch cables that you still can buy are NOT capable of running high speed without problems.

 

Yeah I know, there is no "do not autospeedselect" option in the UNRAID GUI. I usually limit the switche's port to 1G then. But with a "dumb" switch, this is no option too, sorry.

It should be possible with the "ethtool" command, but I am not sure, which syntax is needed.

 

 

Alright, cool. The cables are enterprise grade, gotten off work. There's no doubt they're solid.

Thanks for the help so far, I'll try to look into this and see what I can do and mark solved if I reach the goal.

 

Until then, if anyone else can pitch in with further, any help is appreciated!

 

EDIT: Just remembered. The switch is only 1Gbit ports, so I'm thinking this is all moot and the fault is somewhere else. Or am I off?

Edited March 13, 20233 yr by TalkingWaterfall

1 hour ago, TalkingWaterfall said:

EDIT: Just remembered. The switch is only 1Gbit ports

yeah, you cannot go above 1G with him, so speed is not the problem...

(and cables are not too, even the cheapest stuff can handle 1G today)

 

So I would recommend a more detailed look:

Reboot (to clear the logs)

log into the console

issue a "ping -t somewhere" (like google or so, a server that is known to be "on")

wait until it fails

cancel the ping with CTRL-C

issue "diagnostics" and post it here.

 

Additional test when the error is still active: pull the lan plug, wait a few second, put it back in. Does LAN get reestablished?

 

 

Edited March 13, 20233 yr by MAM59