Opened 18 months ago

Closed 18 months ago

Last modified 18 months ago

#2858 closed task (fixed)

Fix ability to setup old containers like centos 7 on Ubuntu 22.04 hosts

Reported by: robe Owned by: sac@…
Priority: normal Milestone: Sysadmin Contract 2022-II
Component: SysAdmin Keywords:


I was trying to create a container for centos-7 for ci testing on osgeo8, but sadly after upgrading to Ubuntu 22.04 from 20.04 on osgeo8 and osgeo9, I am unable to create a container such as

lxc launch images:centos/7 centtie

Just yields error: Error: The image used by this instance requires a CGroupV1 host system

discussed here -

Doing this works fine on osgeo7 since it's still a Ubuntu 20.04. so I'm going to hold of upgrading osgeo7 and osgeo3 until this issue is resolved, as I'm afraid many of our old debian containers probably relying on cgroupv1 may not start after.

It's unclear at moment if it is a Ubuntu issue or a lxd issue. Other reports state there is no issue with servers that started off as Ubuntu 22.04. It's only an issue with upgraded servers thus suggesting an Ubuntu upgrade issue.

Change History (4)

comment:1 by robe, 18 months ago

Asked on that discussion and was told this, which did the trick on osgeo4.

Edit /etc/default/grub and ensure that GRUB_CMDLINE_LINUX contains systemd.unified_cgroup_hierarchy=false:



Then run:

sudo update-grub

After rebooting I was able to do:

lxc launch images:centos/7 centtie

I'll do the same on the other upgraded servers later tonight since it requires a server reboot.

comment:2 by robe, 18 months ago

I'm going to start run on osgeo8 and osgeo9. I don't expect there to be more than 5 minutes of downtime.

comment:3 by robe, 18 months ago

Resolution: fixed
Status: newclosed

Done also did patch updates on osgeo8 and osgeo9

comment:4 by robe, 18 months ago

Side note: grasswiki on osgeo8 the apache2 service did not start. I'm guessing it's the proxy setting and that it had started before nginx container and so not seeing the proxy, it failed to restart. Manually restarting the apache on it did the trick.

Others seemed to have been okay.

Note: See TracTickets for help on using tickets.