Proxmox | Fix "e1000e Detected Hardware Unit Hang"

fix-eth-hang.md

Proxmox network unavailability problems after upgrading to 8.4.1

After upgrading to 8.x, sometimes access to the proxmox lan/web interface/ssh fails to connect after plugging off and on the ethernet port, it works again for some time and this problem occurs again.

When i connected a monitor and keyboard to the proxmox machine, First i read and thought the new device naming is the problem, so i updated the GRUB_CMDLINE_LINUX field on /etc/default/grub file: it became like this after adding:

GRUB_CMDLINE_LINUX="net.naming-scheme=v252"

After reading i changed the eno1 device name in /etc/network/interfaces file, to some values i've read and restarted the networking service, which made the connection unavailable, and did not worked. Reverting to the original file state and restarting the networking service made the connection returned(just like plugging off and on the cable), but problem occured again later. i left the grub config and did not change it back.

After that i ran dmesg -H i saw Detected Hardware Unit Hang error. This gist is forked from another gist (https://gist.github.com/brunneis/0c27411a8028610117fefbe5fb669d10) and i tried this.

After configuring my machine's /etc/network/interfaces file became like this:

auto lo
iface lo inet loopback


# found this gist and added the post-up scripts - https://shorturl.at/xhX2Y

iface eno1 inet manual
        post-up /usr/bin/logger -p debug -t ifup "(post-up) /sbin/ethtool -K $IFACE gso off gro off tso off tx off rx offrxvlan off txvlan off"
        post-up /sbin/ethtool -K $IFACE gso off gro off tso off tx off rx off rxvlan off txvlan off
        post-up /usr/bin/logger -p debug -t ifup "(post-up) Disabled offload for $IFACE"

auto vmbr0
iface vmbr0 inet static
        address 192.168.1.111/24
        gateway 192.168.1.1
        bridge-ports eno1
        bridge-stp off
        bridge-fd 0


source /etc/network/interfaces.d/*

If problem is gone, i'll update here.

https://pve.proxmox.com/pve-docs/chapter-sysadmin.html#sysadmin_network_configuration

Original gist and the comment that i've used:

Proxmox Virtual Environment 6.1-3 / Debian 10 (buster)
Kernel 5.3.10-1-pve
Ethernet controller: Intel Corporation Ethernet Connection (2) I218-V (rev 05)
Ethernet driver: e1000e 3.2.6-k
Ethernet firmware: 0.1-4

The solution I've found was to create a oneshot service which disables segmentation offloading.

sudo apt update && apt -y install ethtool

Create the file /usr/lib/systemd/system/fix-e1000e.service with the following content:

[Unit]
Description="Fix for ethernet hang errors"

[Service]
Type=oneshot
ExecStart=/usr/sbin/ethtool -K NETWORK_INTERFACE tso off gso off

[Install]
After=network-online.target
Wants=network-online.target

systemctl daemon-reload
systemctl enable fix-e1000e
systemctl start fix-e1000e

This is what i've applied to my machine.

Thanks @CRCinAU

I modified your command to

use multiple "post-up" lines instead of "&&"
use "$IFACE" in the log messages
log the actual ethtool command (since I'm sure I'll forget if I ever see this in the logs...)
disable offloading for rxvlan and txvlan (since I have vlans configured)

iface eno1 inet manual
	post-up /usr/bin/logger -p debug -t ifup "(post-up) /sbin/ethtool -K $IFACE gso off gro off tso off tx off rx off rxvlan off txvlan off"
	post-up /sbin/ethtool -K $IFACE gso off gro off tso off tx off rx off rxvlan off txvlan off
	post-up /usr/bin/logger -p debug -t ifup "(post-up) Disabled offload for $IFACE"