The officially official Devuan Forum!

hoglet

Member

Registered: 2026-01-07

Posts: 3  

Problems when migrating from Debian Trixie to excalibur

I started with migrating from Docker to Podman on Debian, with a small amount of hurdles I got it working.
Then I followed the guide: https://www.devuan.org/os/documentation … libur.html which worked well except that I lost network and had to improvise. I might have misstepped there but I do not see what it would be.

When starting containers I get warnings:

WARN[0000] Using cgroups-v1 which is deprecated in favor of cgroups-v2 with Podman v5 and will be removed in a future version. Set environment variable `PODMAN_IGNORE_CGROUPSV1_WARNING` to hide this warning.

The important containers are running well and fine but I get following in the user.log:

2026-01-08T13:31:59.485264+01:00 master conmon: conmon 2451200db3be06b80bae <nwarn>: Failed to get memory cgroup path. Container may have exited

One interesting part is that elogind is not running (making ssh-login take 25 seconds). Trying to start elogind quietly fails.

When logging in via ssh you get:

2026-01-08T13:38:08.795011+01:00 master dbus-daemon[2543]: [system] Failed to activate service 'org.freedesktop.login1': timed out (service_start_timeout=25000ms)
2026-01-08T13:38:08.795132+01:00 master sshd-session[26143]: pam_elogind(sshd:session): Failed to create session: Failed to activate service 'org.freedesktop.login1': timed out (service_start_timeout=25000ms)

I have a desktop system running freia where elogind starts without a problem and I can start containers without problem. It seems to find cgroup but fails  with:

2026-01-08T13:53:01.166107+01:00 Big conmon: conmon 32f002b0c8f4adfaffbc <nwarn>: Failed to add inotify watch for /sys/fs/cgroup/1/memory.events 
2026-01-08T13:53:01.196039+01:00 Big conmon: conmon 32f002b0c8f4adfaffbc <nwarn>: Failed to open cgroups file: /sys/fs/cgroup/1/memory.events

NOTE: I am running as user on the desktop system and root on server.

So do I have a cgroup problem or a config problem with elogin or something else?

hoglet

Member

Registered: 2026-01-07

Posts: 3  

Re: Problems when migrating from Debian Trixie to excalibur

Update. I found an offending line in /etc/fstab which i commented out:

#cgroup  /sys/fs/cgroup  cgroup  defaults  0   0

So now elogind starts and I do not get the "cgroup warning" when starting containers.

There is still a race condition at boot between a pod and podman creating its socket in /run/podman/podman.sock.
My bind9 pod is not working either so I have to do some more investigation.

hunter0one

Member

Registered: 2021-12-31

Posts: 81  

Re: Problems when migrating from Debian Trixie to excalibur

That line in fstab is what mounts cgroups v2 in the absence of systemd.

Last edited by hunter0one (2026-01-11 16:39:58)

hoglet

Member

Registered: 2026-01-07

Posts: 3  

Re: Problems when migrating from Debian Trixie to excalibur

That makes sense except that I do not run systemd and removing the line makes the system create cgroups in the correct way. So my bet is that that line should not be there.

GlennW

Member

From: Brisbane, Australia

Registered: 2019-07-18

Posts: 696  

Re: Problems when migrating from Debian Trixie to excalibur

The file /etc/rc.conf maybe the file you need to set the config for cgroups.

A quick web search, (Brave-browser) gives all kinds, v1 & v2 as well as hybrid examples.

(I get an error in my boot logs about cpu and mem flags...)

I used the fstab line but my system stalls at startx, it does recover but I generally have given up before then... still learning.

there was a lot more info there...here's the top.

cgroup Mode
The primary setting is rc_cgroup_mode, which determines the cgroups version used:

rc_cgroup_mode="unified" – Uses cgroups v2 only (default in Alpine Linux since v3.19). 
rc_cgroup_mode="legacy" – Uses cgroups v1 only. 
rc_cgroup_mode="hybrid" – Uses both cgroups v1 and v2.
Note: To enable cgroups v2 exclusively, set rc_cgroup_mode="unified" in /etc/rc.conf and reboot for changes to take effect. 

Seems this is related more to openrc service manager. I hope this helps.

---

pic from 1993, new guitar day.