Hi,

When -d is the default, you not only see the problems immediately.
It is much worse: stderr of lxc-start is redirected to /dev/null,
so you don't even see error messages from lxc-start itself!  Such
as, cgroupfs is not mounted, or what not.

I don't think you are making a good argument for not making -d
the default, I think you are making a far better argument for:

 - change lxc-start to print useful errors even with -d
   (especially during initialization)

 - perhaps change lxc-start that if -C is specified that stdin/
   stdout/stderr are then connected to the PTS that is mounted as
   /dev/console within the container, so that the console logfile
   automatically captures the default output
   (instead of using /dev/null, as is done currently)

I think the better alternative, in my opinion, is to be able to
specify -d in the config file.  When you're done, write it there,
and -d becomes the default _for this container_, and you can use
-F to force foreground.

I don't think it's a good idea to modify the daemonizing/non-
daemonizing behavior in the configuration file. This breaks the
expectation that lxc-start $OPTION -n $CONTAINER will do the same
for different containers.

As for the rest, I agree with John's reasoning.

Regards,
Christian

_______________________________________________
lxc-devel mailing list
[email protected]
http://lists.linuxcontainers.org/listinfo/lxc-devel

Reply via email to