El Sun, 20 Apr 2008 18:20:00 +0100
Ian Lee <[EMAIL PROTECTED]> escribió:

> Ricardo Garcia wrote:
> > I don't think (IMHO) that a dependency restart event should be
> > necessary, just dependency stopped, because that's where the
> > restarting problem comes - dependant services always auto-restart
> > when their dependencies go down, so what we want is simply to stop
> > that from happening, right? Unless we do want to take some other
> > action when a dependency restarts. But let's assume that the user
> > wants to shut it down, prepare some stuff and then start it again.
> 
> Thats not restarting(as initng sees it) a service, thats manually 
> stopping and starting a service with two ngc commands
> 
> > For that case, a service should just shut down and restart (which
> > would lead to "WAITING_FOR_DEPS" status) when a dependency is shut
> > down (unless of course the flags discussed in here are set). When it
> > restarts, everything is handled automatically. So, is there really a
> > need for a restart event? Can we find a practical example for it?
> 
> Currently there is a STATE_CHANGE_EVENT whenever the service state 
> changes, there are no restart states. I've been adding these states, 
> which has altered the way initng process restarting events, Iv'e been 
> moving it into the service plugin to follow the start/stop routines
> 
> So a restarting service will be in the states
> 
> RESTART_MARKED -> WAITING_FOR_RESTART_DEPS ->
> SERVICE_RESTART_DEPS_MET ->SERVICE_STOPPED_FOR_RESTARTING
> 
> 
> WAITING_FOR_RESTART_DEPS - is when dependents are stopped (as per
> script variables)
> SERVICE_STOPPED_FOR_RESTARTING  - is when service is restarted
> 

We should be able to decide to propagate the event or not. Most of the
time it's not relevant when a service restarts...

-- 
Ismael Luceno

Attachment: signature.asc
Description: PGP signature

-- 
_______________________________________________
Initng mailing list
[email protected]
http://jw.dyndns.org/mailman/listinfo/initng

Reply via email to