[lxc-users] Unprivileged containers and reboot of base

[Serge Hallyn]

Quoting Medhamsh V (medhamsh at gmail.com):
> Hi,
> 
> Though this was discussed several times I could not get through this. I use
> chef to provision base nodes on which I run unprivileged containers. So
> everything is programmatic and I leave a script which creates and starts
> containers in the base node. I use ubuntu-14.04 ami in Amazon.
> 
> Right after provisioning I cant start the containers unless I reboot the
> machine. The work around for this is to do the following in the sequence
> and after which I can start the containers.
> 
> "sudo cgm create all goatos && sudo cgm chown all user $(id -u user) $(id
> -g user)"

do you mean cgm chown all goatos $(id -u user) $(id -g user) ?

if not then that'll be your problem.

> and then as the unprivileged user,
> 
> cgm movepid all user $$ > /dev/null 2>&1

again, movepid all goatos $$

> Though this solves the problem, upon starting every container I get the
> following messages.
> 
> lxc_container: call to cgmanager_move_pid_abs_sync failed: invalid request
> lxc_container: Failed to enter group /goatos/test
> lxc_container: error communicating with child process
> 
> and everything else is fine. Container starts fine, gets an ip and
> communicates etc.

What is /proc/self/cgroup in one of those containers

> What is the significance of those messages and how to get rid of them?
> 
> Thanks,
> -- 
> Medhamsh,
> 
> GPG Finger Print: BD16 E32E CA4D 83A3 1270  725D D766 7997 0ABC 20E9
> http://pascal.iiit.ac.in/~medhamsh/medhamsh.gpg

> _______________________________________________
> lxc-users mailing list
> lxc-users at lists.linuxcontainers.org
> http://lists.linuxcontainers.org/listinfo/lxc-users