IO limits on LXD with no cgroup inheritance


(Artem) #1

Hi. I try to limit IO for my container. I set limits.write, but it doesn’t work. Systemd in container creates cgroup subdirectory with system.slice and user.slice. Every process gets own subdirectory in one of these slices. Blkio limits defined in /sys/fs/cgroup/blkio/lxc/my_container/blkio.throttle.write_bps_device are not inherit for subdirectories - this is normal behaviour for blkio. The documentation says
“Both CFQ and throttling implement hierarchy support; however,
throttling’s hierarchy support is enabled if “sane_behavior” is
enabled from cgroup side, which currently is a development option and
not publicly available.” (https://www.kernel.org/doc/Documentation/cgroup-v1/blkio-controller.txt).

Is there a some way to force io limits work with systemd in container without enabling “sane_behaviour”?