[Lxc-users] More than one simultaneous lxc-wait?

[Daniel Lezcano]

On 10/05/2010 07:58 AM, Scott Bronson wrote:
> Is it true that only one lxc-wait can happen at a time??
>
> I wrote some scripts that start and stop machines based on the run states of
> other machines (this is for some fairly involved test scenarios).
>   Unfortunately,  only the first script to run manages to lxc-wait -- the
> rest bomb out with "lxc-wait: bind : Address already in use."
>
> Is there a workaround for this?  Trying to rewrite the scripts so they
> negotiate amongst themselves who should next call lxc-wait would be a huge
> undertaking and probably require predicting the future.
>
> Is there any reason the lxc-monitor and lxc-wait manpages don't mention this
> limitation?  I really wish I'd known about it last week!
>
> I found this patch from a year and a half ago that, if related, suggests
> that this is a temporary condition...  Any news?
>
> http://lxc.git.sourceforge.net/git/gitweb.cgi?p=lxc/lxc;a=commitdiff_plain;h=31c53c2e1afc6c81ce4aad5ef4cb982cd2fea365
>    

Right. It was supposed to be temporary :)

We need something like 
http://wiki.virtualsquare.org/wiki/index.php/IPN_examples , but 
unfortunately it is not mainstream.
I looked at extending af_netlink to implement a similar capability in 
the kernel, but I gave out due to a lack of time.

Broadcasting through the loopback interface could be used. I don't like 
this solution because of the port number. But that could be used as a 
better temporary solution :)

Thanks
   -- Daniel