Unit state during charm upgrade

[Kapil Thangavelu]

Excerpts from Gustavo Niemeyer's message of 2012-03-05 20:35:22 -0500:
> On Mon, Mar 5, 2012 at 16:33, Gary Poster <gary.poster at canonical.com> wrote:
> > That was the argument Francesco made, and why he suggested the status name
> > change.
> 
> It would be relatively simple to present the fact a hook is executing
> on a given unit, but at the same time it is expensive to be
> centralizing all that information for every single unit and every
> single hook in the system, just so one can tell if a given script was
> run successfully or not. That's why I'm pondering about how to solve
> the problem without introducing that overhead permanently.
> 
> -- 
> Gustavo Niemeyer
> http://niemeyer.net
> http://niemeyer.net/plus
> http://niemeyer.net/twitter
> http://niemeyer.net/blog
> 
> -- I'm not absolutely sure of anything.
> 

For certain types of hooks (around the unit lifecycle statemachine, ie. non 
relation hooks) we are doing a WAL to zk to allow for recovery/retry if the 
agent dies. What's unclear is what's a good mechanism to expose this for 
interested parties.

cheers,
Kapil