Не создавать иерархию в cpu group

Насколько мне известно, нет:

JoinControllers=cpu,cpuacct net_cls,netprio
Configures controllers that shall be mounted in a single hierarchy. By default, systemd will mount all controllers which are enabled in the kernel in individual hierarchies, with the exception of those listed in this setting.

И, думаю, не будет, т. к. в ближайшем будущем отдельных иерархий вообще не будет.

cgroups = control groups

https://linuxcontainers.org/cgmanager/introduction/

 $ cgm --help
Usage:

cgm ping

cgm create <controller> <cgroup>

cgm chown <controller> <cgroup> uid gid

cgm chmod <controller> <cgroup> mode

cgm chmodfile <controller> <cgroup> file mode

cgm remove <controller> <cgroup> [0|1]

cgm getpidcgroup <controller> pid

cgm getpidcgroupabs <controller> pid

cgm movepid <controller> <cgroup> pid

cgm movepidabs <controller> <cgroup> pid

cgm getvalue <controller> <cgroup> file

cgm setvalue <controller> <cgroup> file value

cgm gettasks <controller> <cgroup>

cgm gettasksrecursive <controller> <cgroup>

cgm listchildren <controller> <cgroup>

cgm removeonempty <controller> <cgroup>

cgm prune <controller> <cgroup>

cgm listcontrollers

cgm listkeys <controller> <cgroup>

cgm apiversion

 Replace '<controller>' with the desired controller, i.e.
 memory, and '<cgroup>' with the desired cgroup, i.e. x1.
 For create, chown, chmod, remove, prune, remove_on_empty,
 gettasksrecursive and movepid, <controller> may be «all» or
 a comma-separated set of cgroups.
 Remove by default is recursive, but adding '0' as the last argument
 will perforn non-recursive deletion.  Adding '1' is supported
 for legacy reasons.

 To refer to the current cgroup, use ".

Мне вообще иерархия не нужна, хочу, чтобы все процессы в /sys/fs/cgroup/cpu/tasks содержались. И никаких дочерних директорий

Результаты ее работы сохраняться после перезагрузки?

хз

Ну а нафига советуешь тогда. Удалять cgroup'ы я умею и с помощью rmdir