Hello, I've found that when lxc-autostart is executed via systemd, the cgroups assigned to the container are wrong for some control groups.
See the following two examples:
1. Container started via systemd (/lib/systemd/system/lxc.service unit
that calls lxc-autostart).
# cat /proc/${pidofsomeprocessinsidethecontainer}/cgroup
9:perf_event:/lxc/clopeztest3
8:blkio:/
7:net_cls,net_prio:/lxc/clopeztest3
6:freezer:/lxc/clopeztest3
5:devices:/
4:memory:/
3:cpu,cpuacct:/
2:cpuset:/lxc/clopeztest3
1:name=systemd:/system.slice/lxc.service
2. Container started manually by first stopping it and then executing
lxc-autostart from a ssh session.
# cat /proc/${pidofsomeprocessinsidethecontainer}/cgroup
9:perf_event:/lxc/clopeztest3
8:blkio:/lxc/clopeztest3
7:net_cls,net_prio:/lxc/clopeztest3
6:freezer:/lxc/clopeztest3
5:devices:/lxc/clopeztest3
4:memory:/lxc/clopeztest3
3:cpu,cpuacct:/lxc/clopeztest3
2:cpuset:/lxc/clopeztest3
1:name=systemd:/user.slice/user-0.slice/session-1.scope
As you can see, when systemd calls lxc-autostart, the cgroups for
memory, devices, cpu, cpuacct and blkio are not correctly set.
The effect of this is that any limitation configured on such container
(for example: amount of memory available for the container) is not
respected.
Are any of you aware of this behavior?
Is there any workaround to fix this?
Regards!
--------
Some notes about the environment:
- This is a Debian 8 (jessie) system running the default LXC version
(lxc=1:1.0.6-6+deb8u2).
- The memory cgroup was enabled (is disabled by default on Debian) by
setting on /etc/default/grub
GRUB_CMDLINE_LINUX_DEFAULT="quiet cgroup_enable=memory swapaccount=1"
and running update-grub && reboot
signature.asc
Description: OpenPGP digital signature
_______________________________________________ lxc-users mailing list lxc-users@lists.linuxcontainers.org http://lists.linuxcontainers.org/listinfo/lxc-users
