[one-users] [RFC] Release 1.4 - Support for hooks

Ruben S. Montero rubensm at dacya.ucm.es
Wed Feb 18 03:03:52 PST 2009


Hi,
	It totally makes sense. It'd be more accurate to say that the event will be 
triggered on entering the state. Otherwise, we'll probably be triggering 
several times the same action (as long as the VM will be in that state).

	It's very interesting your suggestion of possibly assigning different actions 
depending on the original state.  As you said it could be totally different 
enter RUNNING from BOOT or from SUSPEND. From my point of view we have two 
options here:
  * Move the logic to the action. The action it self could check what should 
be performed depending on external conditions (for example if exists a 
checkpoint file or not). Pros: Easy configuration, requires less understanding 
of OpenNebula's state machine. Cons: Actions more difficult to program.

  * Move the logic to OpenNebula core. The user would specify the transition. 
Pros, and Cons the opposite  of above.

As an extra we could automatically add an argument with the source state in 
the transition, so it'd be easier for the action to check what to do.

Thanks for the feedback!


Cheers!

Ruben


 
On Tuesday 17 February 2009 18:15:41 Borja Sotomayor wrote:
> Hi,
>
> >  EVENT: The event that will trigger the execution of the hook. The events
> > are VM states (BOOT,RUNNING...)
>
> Would it make sense for the events to be transitions instead of VM
> states? I'm not familiar with the internal ONE state machine, but if
> there are states with an in-degree > 1, each transition into such a
> state may require a different treatment. For example (and I'm thinking
> of Haizea's state machine), a VM can enter the "RUNNING" state from
> both a "READY" state (the machine has been deployed for the first
> time) and from a "RESUMING" state (the machine is deployed, was
> running, but had to be suspended, and now it's about to start running
> again after resuming), and each transition requires a different
> treatment (even if they both end in the "RUNNING" state). Plus, even
> if all states in ONE currently have an in-degree = 1, reacting to
> transitions instead of states would be a more general solution (in
> case the state machine gets more complicated in the future).
>
> Cheers!

-- 
+---------------------------------------------------------------+
 Dr. Ruben Santiago Montero
 Associate Professor
 Distributed System Architecture Group (http://dsa-research.org)

 URL:    http://dsa-research.org/doku.php?id=people:ruben
 Weblog: http://blog.dsa-research.org/?author=7
             
 GridWay, http://www.gridway.org
 OpenNEbula, http://www.opennebula.org
+---------------------------------------------------------------+





More information about the Users mailing list