Can Debian images be fixed to stop properly?

[deviantintegral (Andrew Berry)]

There’s many postings about Debian containers not stopping correctly, including:

Can't stop debian 8, 9 and 10 containers

It seems to happen on two different hosts: Ubuntu 18.04 (real machine) and CentOS 7 (remote KVM). Both running lxd 3.11 from snapd. If I create debian container: lxc launch images:debian/9 deb and later try to shutdown it, nothing happens, lxc stop deb hangs. However, if I enter into second shell and issue second lxc stop deb command, then first lxc stop deb stops too and container is stopped. It seems, that debian 8, 9 and 10 images are affected, while alpine 3.9 and centos 7 aren’t. Did no…

https://github.com/lxc/lxd/issues/4939#issuecomment-413900839

Masking networkd above is working for me, but it’s a pretty poor evaluator experience for mainline distros to be broken in a basic way like that. Is there some technical reason why the upstream images haven’t been fixed?

[bruce78 (Bruce)]

Getting a fix for this would be great… the old lxc stop --force mycontainer always feels a little odd…

[stgraber (Stéphane Graber)]

stgraber@castiana:~$ lxc launch images:debian/10 debian
Creating debian
Starting debian                             
stgraber@castiana:~$ lxc list
+--------------+---------+----------------------+----------------------------------------------+------------+-----------+
|     NAME     |  STATE  |         IPV4         |                     IPV6                     |    TYPE    | SNAPSHOTS |
+--------------+---------+----------------------+----------------------------------------------+------------+-----------+
| centos3      | STOPPED |                      |                                              | PERSISTENT | 0         |
+--------------+---------+----------------------+----------------------------------------------+------------+-----------+
| centos4      | STOPPED |                      |                                              | PERSISTENT | 0         |
+--------------+---------+----------------------+----------------------------------------------+------------+-----------+
| cosmic       | RUNNING | 10.166.11.42 (eth0)  | 2001:470:b368:4242:216:3eff:fe3e:4813 (eth0) | PERSISTENT | 0         |
+--------------+---------+----------------------+----------------------------------------------+------------+-----------+
| cosmic-cloud | RUNNING | 10.166.11.183 (eth0) | 2001:470:b368:4242:216:3eff:fe5f:5552 (eth0) | PERSISTENT | 0         |
+--------------+---------+----------------------+----------------------------------------------+------------+-----------+
| debian       | RUNNING | 10.166.11.230 (eth0) | 2001:470:b368:4242:216:3eff:fe11:8ac (eth0)  | PERSISTENT | 0         |
+--------------+---------+----------------------+----------------------------------------------+------------+-----------+
stgraber@castiana:~$ lxc stop debian
stgraber@castiana:~$ 

Do you have more details on what’s needed to see such a problem?

[bruce78 (Bruce)]

Ok, I tested lxc stop on a fresh debian 10 image and all was well… I only have this problem on debian 9 images on a debian 9 host but I can live with stop --force!

[stgraber (Stéphane Graber)]

I’ve extended our image testing script to attempt a clean shutdown on all our supported images, so we’ll have a list of broken ones to take a look at soon.

[deviantintegral (Andrew Berry)]

Yes, I can replicate this on the debian 9 image (which I’m using as it’s the recommended version for the Unifi controller app).