Re: monit '-i' option?

[Martin Pala]

Lets discuss whether it is needed then.

I think it could be usefull only in the case, that you start monit from 
init and the services from runlevel rc scripts. In such case there can 
occure race condition between rc script and monit started from init. 
This will not be solved cleanly by skipping validate() in one or more 
monit cycles - such solution only extends the timeframe => in most cases 
the race condition could not happen.

The correct way to solve it is to start such critical services from 
monit directly - monit then will have the service (including start) 
under its control and no false alarms or actions will occure. Btw. this 
way works in real life cluster frameworks too - in the case that you 
need some service to run under cluster's process monitor, you must start 
it via cluster resources.

Though in theory above mentioned solution is clean alternative to -i 
option problem, it is possible that in praxis -i could work well for 
most cases (in the case that the this startup behavior will be 
configurable). In the case that we will readd -i option again, it should 
be well documented including context description (when to use it).

My vote is +0 for readding '-i' with initial test cycles skip count 
configurable extension (as Rory recommended) and -1 for readding it in 
original form (skip validate in first cycle only).

Martin




Rory Toma wrote:

> This should probably be extended to include an integer number.
>
> The syntax would be:
>
> monit -i 3
>
> where 3 would tell how many poll cycles to skip before starting
> monitoring.
>
> Why would you do this? If you start monit from inittab, and you have
> some monitored processes that take can take a long time to come up the
> first time or at boot, it is useful to not try to restart them right
> away.
>
> On Sun, 2003-09-14 at 13:54, Christian Hopp wrote:
>  
>
> > On Sun, 14 Sep 2003, Jan-Henrik Haukeland wrote:
> >
> >    
> >
> > > I didn't add this, and unless someone steps up and explain the
> > > rationale for this switch, I suggest we remove it. The time limit is
> > > two hours :-)
> > >      
> > I have neither used it in the past nor I know what it does!
> >
> >    
> >
> > > Martin Pala <address@hidden> writes:
> > >
> > >      
> > >
> > > > i don't understand validate option '-i'.
> > > >        
> > > [*]
> > >      
> > CHopp
> >    
> >
> > ------------------------------------------------------------------------
> >
> > _______________________________________________
> > monit-dev mailing list
> > address@hidden
> > http://mail.nongnu.org/mailman/listinfo/monit-dev
> >