[SOLVED] Intel NIC e1000e hardware unit hang

lifemoments said:
A helper script is available now -> https://raw.githubusercontent.com/r...fs/heads/main/tools/pve/nic-offloading-fix.sh

Run the following as root.
Code: 
bash -c "$(curl https://raw.githubusercontent.com/rcastley/ProxmoxVED/refs/heads/main/tools/pve/nic-offloading-fix.sh)"

Note : Use only one method - script of manual update of post-up command in /etc/network/interfaces
Click to expand...
After a long period of silence, the damn error reappears. Physically disconnecting (pulling the power cord) brings the affected hosts back on track. Has anyone tried solving the problem by automatically restarting the network after x minutes of no connection (pinging every x minutes) to another, always-reachable host?
I'm currently trying your great script and will report back.
 
  • Like
Reactions: EnRav
Now that's what I call an IP KVM with a decent price. Have you tried it yet?
Thanks for the tip.
 
Impact said:
Wouldn't buying a proper functioning USB NIC have been cheaper?
Click to expand...
Even though it creeps me out, such a "solution" is increasingly becoming my focus. It's simply annoying to squash software problems with hardware. Unfortunately, Intel doesn't seem to be able to get the problems of its bread-and-butter network chip under control, or at least they simply don't care.
 
TErxleben said:
Even though it creeps me out, such a "solution" is increasingly becoming my focus. It's simply annoying to squash software problems with hardware. Unfortunately, Intel doesn't seem to be able to get the problems of its bread-and-butter network chip under control, or at least they simply don't care.
Click to expand...
For me is the USB dongle certainly not the solution. Most USB device drivers are full of bugs. People just don't pay much attention to it as it's just pull/plug in back and it works again thing.
 
TErxleben said:
Now that's what I call an IP KVM with a decent price. Have you tried it yet?
Thanks for the tip.
Click to expand...
Not yet. It arrived just today and I had not even time to unbox. I did not buy the pin out adapter yet. I may in the future if needed.
 
In my opinion, USB network cards are more of an additional problem than a solution. I've already ordered your idea. I'm curious to see what works. I'll report back. But it's annoying to have to mess around like this.
 
Last edited:
As promised, a quick update. The GL-KVM works quite well, even if the web GUI has a few annoying quirks. The price/performance ratio is really good, though.
 
Impact said:
a proper functioning USB NIC
Click to expand...

As someone who has tried a good number of USB NICs because I have reason to be running Proxmox on physically very small hardware, I can assure you that USB NICs without a ton of bugs are about as common as unicorns. And I really really wanted that not to be the case. But sadly, it is the case.
 
Not to partially necro and/or guide the discussion down a tangent, but I'm really curious - what led to this bug? I've been running Proxmox on an Intel NUC for years and this only reared it's head in the last few months. Not saying Proxmox did something wrong, but what bug/fault/error/update caused this to start to be a problem now, when it wasn't before? Genuinely curious here - is it a Linux kernel thing or something?
 
Can confirm adding this to interfaces helps with my onboard intel I219-V (rev 10) nic on Proxmox 9.0.5. No problems for a week.

Code: 
post-up ethtool -K eno1 gso off tso off rxvlan off txvlan off gro off tx off rx off sg off

Before that, my nic crashed multiple times per day, especially under load, but not necessarily.

Thank you for the solution!
 
Last edited:
Ran into this issue while backing up some VMs. i thought it was Veeam but it could replicate it with PBS. Wild this is still an issue.
 
Just a tip, you only need tso off.
There's no need to turn off all others.
I've been using this on /etc/network/interfaces for over 2 years with no issues:

Code: 
post-up ethtool -K eno1 tso off
Code: 
post-up ethtool -K vmbr0 tso off
 
  • Like
Reactions: thulium and hmohr
Ok, well, that didn't take long. Still have hardware hangs. If I go onto a VM and run:
Code: 
iperf3 -c 192.168.1.148 -P 4 -t 3600
it just crashes.
Latest version of proxmox. Not sure what to do. Tried numerous fixes listed here already, no luck. But, I haven't tried them all. What next?
EDIT:
Guys I can honestly not decide what to think of this. This occurs almost instantly when running a windows VM & IPERF3 with a VirtIO "NIC"
View attachment 91439
Code: 
Oct 07 22:41:41 proxmox kernel: e1000e 0000:00:1f.6 eno1: Detected Hardware Unit Hang:
                                  TDH                  <e3>
                                  TDT                  <13>
                                  next_to_use          <13>
                                  next_to_clean        <e2>
                                buffer_info[next_to_clean]:
                                  time_stamp           <1005c1c81>
                                  next_to_watch        <e3>
                                  jiffies              <1005c9600>
                                  next_to_watch.status <0>
                                MAC Status             <80083>
                                PHY Status             <796d>
                                PHY 1000BASE-T Status  <3800>
                                PHY Extended Status    <3000>
                                PCI Status             <10>
However, a linux machine with the "e1000e network device" can run for hours with no errors:
View attachment 91440
go figure.
 
Last edited:
So I'm planning to rebuild my Home Assistant server over the weekend, using Proxmox so that I can build HAOS in a VM and be considered "supported" by HA (today I run HA supervised, but it's fallen out of favour for prod builds). My HA server is an Intel NUC (NUC8I3BEH 8th Gen Core i3). The NIC shows up as:

Code: 
% lspci -v | grep Ethernet
00:1f.6 Ethernet controller: Intel Corporation Ethernet Connection (6) I219-V (rev 30)
    Subsystem: Intel Corporation Ethernet Connection (6) I219-V

I assume, therefore, that I am going to join the club of e1000e driver hangs, which doesn't fill me with joy (the NIC has been 100% stable running bare metal for the last 6 years).

My question: can I avoid this grief if I use the VirtIO network driver instead of Intel E1000? I'm not doing anything fancy with networking - no VLANs or weird IP config. I'll just have separate IP addresses for the host and each of the VM guests. I do have Gbps fibre from my ISP, so would like to maintain full Gbps performance.
 
  • Like
Reactions: mr.hollywood
You must log in or register to reply here.
Top Bottom