UI inaccessible after a seemingly random amount of time

[sardaukar]

I've installed Proxmox VE on a Beelink EQ12 mini-PC (16GB RAM, 2x Intel 226-V, AX201 Wifi) and it works great. I have 2 bridges setup for an OPNSense VM to use the physical NICs, and added a TP-Link USB-C to Ethernet 1Gb adapter (ASIX AX88179) to use as the management connection. And it works fine, until it just stops working. The OPNSense VM continues to work fine, because I still have internet access, but not management via Wi-Fi or the USB-C dongle.

What I've tried:

- cron job to `ifreload -a -X enp1s0 -X enp2s0 -X lan -X wan` every 5 minutes (lan and wan are what I call the bridges I created for the OPNSense VM)
- adding `pcie_port_pm=off pcie_aspm=off` to the kernel command line

Nothing works, I keep getting locked out of the UI after a few hours. Curiously enough, both the USB-C dongle _and_ the WiFi connection fail at the same time.

Now I've added a new cronjob every five minutes -> `ping google.com || (ip link set wlo1 down ; ip link set wlo1 up)` (one for each interface, this is for the wi-fi one) hopefully that will fix it but it's weird I need to do this.

I'm sure it's this particular PC, since I have Proxmox running on another non-mini PC and it's rock solid, no network drops.

Here's my `/etc/network/interfaces`:

auto lo
iface lo inet loopback

auto enxf8e43beb440b
iface enxf8e43beb440b inet manual

auto enp1s0
iface enp1s0 inet manual

auto enp2s0
iface enp2s0 inet manual

auto vmbr0
iface vmbr0 inet static
        address 192.168.2.2/20
        gateway 192.168.2.1
        bridge-ports enxf8e43beb440b
        bridge-stp off
        bridge-fd 0

auto wlo1
iface wlo1 inet manual

auto lan
iface lan inet manual
        bridge-ports enp1s0
        bridge-stp off
        bridge-fd 0

auto wan
iface wan inet manual
        bridge-ports enp2s0
        bridge-stp off
        bridge-fd 0

source /etc/network/interfaces.d/*

I've tried changing `wlo1` to static with an IP of `192.168.2.9/20` and gateway of `192.168.2.1`, but then I lose access to the UI immediately. The gateway IP is the LAN address of the OPNSense VM. Both the dongle and wifi interfaces get their IP by DHCP, which runs on the OPNSense VM as well.

Any pointers much appreciated!

 

[Moayad]

Hi,

Did you see anything related to the network in `dmesg`?

 

[sardaukar]

Just the usual `entered blocking / disabled state` messages that show up a lot on the logs

 

[hyperXL99]

Lol I have exactly the same issue. Does the cronjob ping "fix" the problem for you?

 

[sardaukar]

I've replaced the cronjob with `ip link set $INTERFACE down; ip link set $INTERFACE up` every 5 minutes and it doesn't get inacessible anymore.

Kinda ridiculous that Proxmox has this issue for some hardware configurations, really puts me off getting a subscription if their product is this brittle (well, that and the cost).

 

[hyperXL99]

I've replaced the cronjob with `ip link set $INTERFACE down; ip link set $INTERFACE up` every 5 minutes and it doesn't get inacessible anymore.

Kinda ridiculous that Proxmox has this issue for some hardware configurations, really puts me off getting a subscription if their product is this brittle (well, that and the cost).

So you deleted the ping part of the cronjob? I tried this now and hope that it works.

But I also would like to know what causes this.

 

[sardaukar]

Yeah I got rid of the ping because I couldn't make it ping from a specific interface

 

[hyperXL99]

Update: It doesn' t work

*/5 * * * * /usr/sbin/ip link set enp5s0 down && /usr/sbin/ip link set enp5s0 up

Still was offline at 4 in the morning. Only fix was typing systemctl restart networking using an attached keyboard.

 

[sardaukar]

I have mine as (root's crontab):

*/5 * * * * ip link set wlo1 down ; ip link set wlo1 up
*/5 * * * * ip link set enxf8e43beb440b down ; ip link set enxf8e43beb440b up

And it's been working fine for 3 days now

 

[hyperXL99]

Yeah mine is also a root crontab but it just doesn't work. Interestingly connections from pve to outside work just fine (Grafana monitoring continuously gets new data), it's just that I cant access it from outside. It would be nice it anyone from the dev team would have a look at this.

 

[sardaukar]

Ok spoke too soon - it just went down for me after DAYS. This is so random :/