Incus fails to allow forwarded ports on Void due to systemd expectations....

[acidvegas]

Is this a new report?

Yes

System Info

Void

Package(s) Affected

incus

Does a report exist for this bug with the project's home (upstream) and/or another distro?

lxc/incus#625

tldr; the solution is to run:

mkdir /sys/fs/cgroup/systemd
mount -t cgroup -o none,name=systemd systemd /sys/fs/cgroup/systemd 

and then you can start your containers......I guess incus on void is expecting systemd mounts in the cgroup

Classic case of devs only making things catered to ubuntu/debian builds

Reporting this here, doubt anyone is going to fix it.

Cheers

Expected behaviour

Containers running

Actual behaviour

Containers wont run and bind to forwarded ports.

Steps to reproduce

read the linked issue

[ahesford]

Try changing CGROUP_MODE to unified in /etc/rc.conf, reboot, and see if the problem goes away without the extra mount.

[acidvegas]

Try changing CGROUP_MODE to unified in /etc/rc.conf, reboot, and see if the problem goes away without the extra mount.

no dice.

[brandon@paloaltonetworks ~]$ incus start elasticsearch-container
Error: Error occurred when starting proxy device: Error: No such file or directory - Failed to safely open namespace file descriptor based on pidfd 3

Same error as I was getting in the issue I linked.

[dkwo]

How exactly can I reproduce this?
With unified setup, I can run e.g. Ubuntu containers (without the mkdir/mount thing that was also in lxd runit service)

$ incus launch images:ubuntu/24.04 u1t
Launching u1t
$ incus list
+------+---------+--------------------+------+-----------+-----------+
| NAME |  STATE  |        IPV4        | IPV6 |   TYPE    | SNAPSHOTS |
+------+---------+--------------------+------+-----------+-----------+
| u1t  | RUNNING | 10.60.1.240 (eth0) |      | CONTAINER | 0         |
+------+---------+--------------------+------+-----------+-----------+

[classabbyamp]

if i had to guess this cgroup mount is done by elogind