Today I noticed that none of my containers were getting IP addresses.
I found the following message in lxd.log:
err="readlink /proc/591/exe: no such file or directory" lvl=eror msg="Failed to bring up network" name=lxdbr0 t=2017-06-19T14:03:36+1200
PID 591:
$ ps 591
PID TTY STAT TIME COMMAND
591 ? S< 0:00 [loop5]
$ sudo ls -l /proc/591/exe
ls: cannot read symbolic link '/proc/591/exe': No such file or directory
lrwxrwxrwx 1 root root 0 Jun 15 10:02 /proc/591/exe
$ _
After some searching around I found https://github.com/lxc/lxd/issues/2767 which suggested a problem with the dnsmasq pid file, and indeed:
$ cat /var/lib/lxd/networks/lxdbr0/dnsmasq.pid
591
$ _
I deleted this file and restarted the lxd service and my containers shortly received IP addresses. This file was probably became stale thanks to a recent hard reboot of the host machine. It would be best if LXD could recover from this condition itself somehow, or at least provide a hint to the operator, as it seems non-trivial to debug, but storing this piece of state in a location that does not survive a reboot might also work.
What version of LXD is that?