Re: combining automatic state transistions with actioned transistions OK?

["marios@redhat.com" <ma...@redhat.com>]

On 02/03/12 02:46, Koper, Dies wrote:
> Please review this subset of transitions for FGCP. Is the
> pending->:finish valid here?
> 
>       start.to( :pending )          .on( :create )  # new instances do
> not start automatically
>       pending.to( :stopped )        .automatically  # after creation
> they are in a stopped state
>       pending.to( :finish )         .on(:destroy)   # after creation
> they can be destroyed
>       stopped.to(:finish)           .on( :destroy ) # only destroy
> removes an instance, and it has to be stopped first
> 

see my reply to your other e-mail here... it is hard to gage whether
this is 'correct' - this depends entirely on FGCP. My advice is to get
the driver out and wait for comments from other FGCP users.

> If 'pending' transitions to :stopped automatically, the user wouldn't
> have a chance to action :destroy on it, so I suppose only the fourth
> transition is possible. Should I remove pending->:finish?
> 
> What confused me is the following in ec2 driver:
> 
>           pending.to( :running )        .automatically
>           pending.to( :stopping )       .on( :stop )
>           pending.to( :stopped )        .automatically
> 
> How can a state automatically transition to two opposite states, and at

yes, the ec2 driver is a special case because EC2 supports both
stateless (instance-store backed) and stateful (ebs backed) instances;
there is an ongoing discussion as to how to handle this case.

marios

> the same time allowing a user to action on it?
> 
> Thanks,
> Dies
>