Very odd container networking issue

LXD
1

For 1x container, always the same LAMP stack site running in the container, periodically if appears lxdbr0 stops routing packets.

Pointing mtr to the broken container show packet flow all the way to the container, then there’s simply no response.

Stopping the container. Changing IP. Restarting container. Problem is fixed.

And, even waiting days, assigning the old IP to a new container, problem persists.

Likely this problem can be fixed via a reboot + this machine is running production sites + takes a very long time to boot.

Someone let me know if there’s some way to debug this or maybe to bounce (stop + restart) LXD networking to 1x container or all containers.

So a few seconds of outage across all containers is fine.

I’d just like to avoid a reboot to fix this.

Thanks.

2

Anything useful in dmesg?
I guess it could be some kind of MAC/IP conflict or the ARP table being full?

3

IPs involved are 144.217.229.110 + 144.217.229.111 which are both still dead, for several days.

http://144.217.229.99/dmesg.txt provides the output of dmesg.

If you notice anything odd + have a fix to try, let me know.

Thanks.