[lxc-devel] [patch 1/1] fix container find the previously created configuration
Daniel Lezcano
dlezcano at fr.ibm.com
Fri Nov 20 13:38:50 UTC 2009
Andrian Nord wrote:
> On Thu, Nov 19, 2009 at 10:33:35AM +0100, Daniel Lezcano wrote:
>> lxc-wait is broken ?
>
> I'm starting container's in a threaded way (many at once), but if i'll
> use say lxc-wait -n <name> -s RUNNING (I want to known, when container
> is actually started, to prevent hundreds on containers starting at once,
> which will only slowdown boot process), one will lock PF socket and
> other's (lxc-monitor also) would fail with 'address in use' error.
>
> hellgate night # lxc-wait -n nfs -s STOPPED
> <waiting...>
>
> in other console:
> hellgate night # lxc-wait -n openafs -s STOPPED
> lxc-wait: bind : Address already in use
>
> Same thing with lxc-monitor.
>
> Solution I see (and i'm planning to implement it one day) in implementing
> 'daemon' mode for lxc-monitor and modifying monitor.c to check if
> monitor-daemon is running before (it could be forced with special flag
> for lxc-monitor/lxc-wait and will issue error, if no lxc-monitor-daemon
> running) trying to direct access. lxc-monitor in daemon mode will lock
> socket and create normal af_unix socket, redirecting all info it got from pf
> socket.
I had in mind to implement in the kernel an AF_IPC or AF_IPN to allow
different processes to communicate together via a message passing
mechanism in the kernel.
This is something similar to http://lwn.net/Articles/261270/ but on top
of the netlink.
More information about the lxc-devel
mailing list