Proxmox 6.8.12-9-pve kernel has introduced a problem with e1000e Driver and network connection lost after some hours

Using the documented workarounds my test box has been up for 21 days with no network drops.

In /etc/networks/interfaces:

Code: 
iface eno2 inet manual
        post-up ethtool -K eno2 tso off gso off
 
  • Like
Reactions: chrisn-au
I am running PVE 9.2.3 for few weeks now. It uses the Linux 7.0.6-2-pve (2026-05-29T11:08Z) kernel and I had no issues so far. But also I still have the offload turned off and the reset script running.
If you need more detail, I can provide that, but need the step-by-step commands as I am not a linux pro.
 
There are so many updates these days that require a reboot that I simply can't say how long the system can run without a freeze. So far so good, the script did not have to kick in. (no log entries yet)

Bash: 
Install ethtool
apt install ethtool


replace aaa.bbb.ccc.ddd with router's IP address
Save it to a file, e.g., /usr/local/bin/network_watchdog.sh
Make it executable
    chmod +x /usr/local/bin/network_watchdog.sh
run:
    crontab -e
Add the line:
    */5 * * * * /usr/local/bin/network_watchdog.sh


NOTE: no need for restart, crontab picks up the new config automatically


Driver detection TEST:
ethtool -i $(for i in /sys/class/net/*/; do i=$(basename "$i"); [[ "$(ethtool -i "$i" 2>/dev/null | awk '/^driver:/{print $2}')" == "e1000e" ]] && echo "$i" && break; done)

=======================================================
#!/bin/bash
TARGET_IP="aaa.bbb.ccc.ddd"
DRIVER="e1000e"
myDateFile="/tmp/network_watchdog_date"
lockFile="/tmp/network_watchdog.lock"
logFile="/var/log/check_network.log"


exec 9>"$lockFile"
/usr/bin/flock -n 9 || exit 0


IFACE=$(for i in /sys/class/net/*/; do
  i=$(basename "$i")
  [[ "$(/usr/sbin/ethtool -i "$i" 2>/dev/null | /usr/bin/awk '/^driver:/{print $2}')" == "$DRIVER" ]] && echo "$i" && break
done)


ret=$(/usr/bin/ping -c3 -W2 -q $TARGET_IP >/dev/null; /usr/bin/echo $?)
if [[ "0" != "$ret" ]]; then
  sleep 30
  ret=$(/usr/bin/ping -c3 -W2 -q $TARGET_IP >/dev/null; /usr/bin/echo $?)
fi


if [[ "0" != "$ret" ]]; then
  curDate=$(/usr/bin/date +"%d.%m.%Y %H:%M")
  if [[ ! -f $myDateFile ]]; then
    echo "$curDate" > $myDateFile
    echo "$curDate: Fault detected on $IFACE, restarting networking" >>$logFile
    /usr/bin/systemctl restart networking
  else
    echo "$curDate: Fault persists on $IFACE, reloading $DRIVER driver" >>$logFile
    /usr/sbin/rmmod $DRIVER
    sleep 2
    /usr/sbin/modprobe $DRIVER
    sleep 5
    /usr/bin/systemctl restart networking
  fi
else
  if [[ -f $myDateFile ]]; then
    myDate=$(cat $myDateFile)
    curDate=$(/usr/bin/date +"%d.%m.%Y %H:%M")
    echo "$curDate: Recovered from fault detected on $myDate" >>$logFile
    rm -f $myDateFile
  fi
fi


=======================================================
 
You must log in or register to reply here.
Top Bottom
Back