[sheepdog] Exit status of node recovery

[Hitoshi Mitake]

At Mon, 26 Aug 2013 13:24:08 +0800,
Liu Yuan wrote:
> 
> On Fri, Aug 23, 2013 at 05:53:02PM +0200, Valerio Pachera wrote:
> > Hi, I'm wondering if it could be good to have an exit status different
> > from 0 when there's a recovery running.
> > 
> > I wish to monitor if cluster is doing a recovery in nagios/icinga or
> > another application.
> > 
> > To get this info I have to "parse" the output of 'node recovery'
> > instead of the exit status.
> > 
> >   lines=$(dog vdi recovery | wc -l)
> >   if [ lines -ne 2 ]; then
> >       echo "recovery is running"
> >   fi
> > 
> > Instead of
> > 
> >   dog vdi recovery > /dev/null
> >   if [ $? -eq 1 ]; then
> >       echo "recovery is running"
> >   fi
> > 
> > What do you think?
> 
> I don't think first form is better than the second form. We use non-zero status
> code to indicate error of executing request as a general rule.

IIUC, non-zero values which indicate error are only EXIT_FAILURE (1)
and EXIT_SYSFAIL (2). Other code in include/exits.h is not for simple
errors. Using a new value for indicating recovery process can be
accepted.

Thanks,
Hitoshi