[lxc-users] is memory.limit_in_bytes inherited by nested cgroups?

[Harald Dunkel]

Hi folks,

I have to restrict lxc.cgroup.memory.limit_in_bytes to 16GByte
for the containers. Problem: New systems based on Stretch show 

% for i in $(find /sys/fs/cgroup/memory/lxc/lxc1 -name memory.limit_in_bytes); do \
	echo $i $(cat $i) \
done | column -t
/sys/fs/cgroup/memory/lxc/lxc1/memory.limit_in_bytes               17179869184
/sys/fs/cgroup/memory/lxc/lxc1/user.slice/memory.limit_in_bytes    9223372036854771712
/sys/fs/cgroup/memory/lxc/lxc1/init.scope/memory.limit_in_bytes    9223372036854771712
/sys/fs/cgroup/memory/lxc/lxc1/system.slice/memory.limit_in_bytes  9223372036854771712


Does this mean that the nested memory cgroups are not 
restricted? I had hoped there is some inheritance in place,
making it impossible for the containers to override memory
restrictions. Does it?


Every helpful comment is highly appreciated
Harri