The init script that comes with 0.8.0-rc1 assumes the cgroup path to be: /sys/fs/cgroup/cpuset/lxc/ which causes the init script to say the starting of the container fails even if it's up and running and the init script can't stop the container at all, as it don't think it's up and running. As the cgroup path can be different depending on the modules built for the host kernel, then the path shouldn't be hard coded in the init script, but be configurable in a /etc/conf.d/lxc config file, so that those who have the cgroup path of /sys/fs/cgroup/lxc/ don't have to worry that at the next update of lxc they have to modify the init script again. Even better had been if the init script could find the correct path without configuration and use it, but I guess that would require a lot more work and knowledge of all the possible paths that are in use and may be use in the future, so the configuration file would be more than acceptable.
Created attachment 312285 [details] Kernel configuration which makes the cgroup directory to be /sys/fs/cgroup/lxc/
What is the configuration you're attaching? It's definitely not text/plain but I'd rather have just the settings for CGROUPS.
Sorry, the config.gz is from my desktop machine, not from the "server". Here is the CGROUPS set in the "servers" kernel config: # zgrep CGROUP /proc/config.gz CONFIG_CGROUPS=y # CONFIG_CGROUP_DEBUG is not set CONFIG_CGROUP_FREEZER=y CONFIG_CGROUP_DEVICE=y CONFIG_CGROUP_CPUACCT=y CONFIG_CGROUP_MEM_RES_CTLR=y CONFIG_CGROUP_MEM_RES_CTLR_SWAP=y CONFIG_CGROUP_MEM_RES_CTLR_SWAP_ENABLED=y # CONFIG_CGROUP_MEM_RES_CTLR_KMEM is not set # CONFIG_CGROUP_PERF is not set CONFIG_CGROUP_SCHED=y CONFIG_BLK_CGROUP=y # CONFIG_DEBUG_BLK_CGROUP is not set # CONFIG_NET_CLS_CGROUP is not set CONFIG_NETPRIO_CGROUP=m
Should have mentioned which kernel too: # cat /proc/version Linux version 3.3.4-gentoo (root@nfs) (gcc version 4.5.3 (Gentoo 4.5.3-r1 p1.0, pie-0.4.5) ) #1 SMP Tue May 8 22:05:37 CEST 2012
I do understand that this has not been fixed in the init script but maybe this is not an issue anymore with recent kernels? I do have /sys/fs/cgroups/cpuset so maybe the path has been standardized in the recent kernels? Please re-open if you still have problems