Using systemd 232, /sys/fs/cgroup/systemd is using cgroup2, causing lxc-start images to fail due cgroup permission (Operation not permitted) There is a workaround for that? I tried https://github.com/debops/ansible-lxc/issues/15#issuecomment-87253770. I know this seems to be more a bug between systemd or lxc userspace tools, but I would like to keep this bug here as existing issue and if someone can let some workaround meanwhile and/or use it as base to report properly to systemd or lxc userspace tools with more technical domain of this matter. I can confirm this works fine with systemd 231, and systemd loading cgroup v1 for systemd. 232 uses cgroup2 causing the problem.
As additional info, using lxc on Ubuntu images, xenial amd64
I do not use lxc, and I am unfamiliar with its use of cgroups. You should work with upstream (systemd and lxc) directly on this.
*** Bug 599268 has been marked as a duplicate of this bug. ***
This has been reverted upstream. We will pick it up with systemd-233, or whenever I backport patches.
This has been fixed here: https://github.com/systemd/systemd-stable/commit/fb36bef4e4884a62b70cd98f7d2d52abf8091106 As a workaround, just append this to your GRUB_CMDLINE_LINUX_DEFAULT: systemd.legacy_systemd_cgroup_controller=yes So it looks something like: # /etc/default/grub GRUB_CMDLINE_LINUX_DEFAULT="quiet systemd.legacy_systemd_cgroup_controller=yes" This is affecting docker as well (irrelevant to this ticket, I know).
Maybe this backport would be useful for us: https://patchwork.openembedded.org/patch/134978/
this should be fixed in 233 finally