Open CalvoM opened 11 months ago
I've been experiencing the same behavior since upgrading to Kubuntu 23.10 for Wired connection 2
being deactivated.
I've verified that I don't experience actual network disconnectivity, just an annoying notification that I have silenced.
@tomponline Is there any progress on this? I am still experiencing this.
Hi @CalvoM we not yet looked at this case. I've asked @gabrielmougard to see if he can reproduce it and diagnose the issue. Thanks
I have tried monitoring the lxdbr0
status (every 0.1s) to check for any bridge flapping while restarting one of my container (u1
), and here is what I got:
In my case, this is very brief (I actually never noticed that before). I guess that depending on your host NIC, this delay can vary. I'll investigate this.
@mr-cal @CalvoM @tomponline I found a hack:
In order to ensure that the bridge is always up, regardless of whether containers are running, I created a dummy interface on the host and attached it to lxdbr0
to keep it "up" even when no containers are running:
ip link add dummy0 type dummy
ip link set dummy0 master lxdbr0
ip link set dummy0 up
...which should keep the bridge from going down when the last container is stopped. Here is my result:
My flapping seems to have disappeared. Can you try that on your side to see if this is related? If yes, I'll work on a LXD fix to setup this behaviour internally.
I think I might be having the same or at least a very similar issue on my setup. I am running a Dietpi OS setup. (For those unfamiliar, Dietpi is a minimal fork of Debian, with focus on optimisations for single board computers such as Raspberry Pi).
I initially posted this issue on Dietpi's issue tracker https://github.com/MichaIng/DietPi/issues/6884, but after some investigation, the issue seems to be with lxd itself rather than with Dietpi.
On my system the hosts network connection is lost "permanently", until it is manually brought online with sudo ifup eth0
, or after a reboot. However my setup is headless server setup, which might be making the difference?
An interesting thing is that after the network interface as manually been brought up with sudo ifup eth0
or even sudo ifup --alow=hotplug eth0
, the network interface is stable until next reboot, or until sudo systemctl daemon-reload
is run.
This means that subsequent container stops and restart does not bring down the network interface, until the next reboot or systemd deamon-reload.
As a workaround, I have found that modifying the /etc/network/interfaces
file, so that the primary network interface is set to auto eth0
instead of allow-hotplug eth0
, which is default on Dietpi as well as on Debian Bookworm, solves the issue.
The issue is not present in a fresh headless install of Debian Bookworm, even though that Debian and Dietpi seems to share the exact same network setup. That is to say that on Debian, is the issue not present even though the network interface is configured with allow-hotplug eth0
.
@CalvoM can you reproduce this workaround ?
Required information
Issue description
Whenever I restart or stop a container my network is interrupted and I lose connection for a second.
Steps to reproduce
Information to attach
Here is the output of
ip a
Output of 'ip -r`
Output of
lxc network show lxdbr0
Output of
lxc config show SampleJammy --expanded
dmesg
)lxc info NAME --show-log
)lxc config show NAME --expanded
)lxc monitor
while reproducing the issue)