LXD crash ? (3.15)


On several host, I can see that lxd process just disapear (crash ?), since I upgraded to 3.15.
About 10% of the host have a lxd crash/day.

Part of lxd.log:

t=2019-07-22T10:45:47+0000 lvl=info msg=“Pruning expired container backups”
t=2019-07-22T10:45:47+0000 lvl=info msg=“Done pruning expired container backups”
[lxd crash here, manual restart]
t=2019-07-22T12:29:16+0000 lvl=info msg=“LXD 3.15 is starting in normal mode” path=/var/lib/lxd
t=2019-07-22T12:29:16+0000 lvl=info msg=“Kernel uid/gid map:”
t=2019-07-22T12:29:16+0000 lvl=info msg=" - u 0 0 4294967295"
t=2019-07-22T12:29:16+0000 lvl=info msg=" - g 0 0 4294967295"

Anyone having this ?
How could I better understand the problem ?

Is there any information about the crash making it to that log file or to the systemd journal or console?

I can not see anything in logs.
Keep having crashes randomly :confused:

I tried to start lxd in debug mode (-d) , but it completely freeze it, lxc command to not work anymore (lxc list blabla for example stay freezed).
Also unable to kill lxd process in debug mode, had to kill -9 .

I’d try something like:

  1. Stop the current lxd process, for example if you are using systemd:
systemctl stop lxd
  1. Launch the lxd process by hand from a console:
sudo lxd --verbose --debug --group lxd

and use lxd normally. Once it crashes, please send us the full output from the console.

Believe it or not, since we restarted all lxd daemons in debug/verbose mode, I didn’t see any crash.

Well, that’s frustrating :slight_smile:

Also odd because we don’t actually skip code when not in debug mode. We actually effectively always run in debug mode and only route error messages differently based on those command line options.
That’s why lxc monitor --pretty --type=logging weill get you a live debug log even when the daemon wasn’t started in debug mode.