destroy has to include stopping? how to not support reboot?

["Koper, Dies" <di...@fast.au.fujitsu.com>]

The API documentation has:

destroy           Stop the instance and completely destroy it

FGCP's destroy equivalent does not stop it first. I first need to stop
it, wait until it's stopped, then destroy it (may even have to detach
resources such as global IP addresses from it first).

Is it enough to just NOT include the following state transition, or do I
need to do anything more?
(I see only GoGrid and SBC have it)

    running.to( :finish )       .on( :destroy )

Similar question for reboot:
FGCP doesn't have a single operation for it. I first need to stop it,
wait until it's stopped, then start it again.

Is it enough to make the reboot_instance method do:

      raise "Reboot action not supported"

and NOT include the following state transition, or do I need to do
anything more?

      running.to( :running )        .on( :reboot )

I noticed condor driver has this state instance but reboot_instance is
not implemented??
I also noticed gogrid driver has reboot_instance implemented but this
state transition is not included??

Cheers,
Dies

Re: destroy has to include stopping? how to not support reboot?

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

Hi Dies:

On 02/03/12 02:31, Koper, Dies wrote:
> The API documentation has:
> 
> destroy           Stop the instance and completely destroy it
> 
> FGCP's destroy equivalent does not stop it first. I first need to stop
> it, wait until it's stopped, then destroy it (may even have to detach
> resources such as global IP addresses from it first).
> 
> Is it enough to just NOT include the following state transition, or do I
> need to do anything more?
> (I see only GoGrid and SBC have it)

there is no 'right' vs 'wrong' with respect to instance states - the
instance life-cycle is entirely dictated by the given cloud provider.

I don't know enough about the FGCP cloud - however, from what you
describe above - there is a 'destroy' action which completely stops the
instance and frees resources, and there is also a 'stop' action which
just stops the instance (pull the plug vs graceful shutdown?)

running.to(:finish) .on(:destroy)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
i.e. invoke the kill+free resources functionality (i.e. 'pull the plug'?)


running.to(:shutting_down) .on(:stop)
shutting_down.to(:stopped) .automatically
stopped.to(:finish) .on(:destroy)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
i.e. invoke the stop. (I assume an intermediate 'pending' state while it
is gracefully shutdown). Then from the stopped state it can also be
destroyed to kill+free resources.


> 
>     running.to( :finish )       .on( :destroy )
> 
> Similar question for reboot:
> FGCP doesn't have a single operation for it. I first need to stop it,
> wait until it's stopped, then start it again.
> 

so you don't *have* to have a reboot action and method, if this is not
supported by FGCP (from a quick look I see the rhevm driver doesn't do
reboot). So combined with the above you could do something like:

running.to(:shutting_down) .on(:stop)
shutting_down.to(:stopped) .automatically
stopped.to(:finish) .on(:destroy)
stopped.to(:pending) .on(:start)
pending.to(:running) .automatically


marios

> Is it enough to make the reboot_instance method do:
> 
>       raise "Reboot action not supported"
> 
> and NOT include the following state transition, or do I need to do
> anything more?
> 
>       running.to( :running )        .on( :reboot )
> 
> I noticed condor driver has this state instance but reboot_instance is
> not implemented??
> I also noticed gogrid driver has reboot_instance implemented but this
> state transition is not included??
> 
> Cheers,
> Dies
>