[Haizea] Conceptual question

[Nikola Milutinovic]

I have one conceptual question.

If I understand correctly, Haizea doesn't like when you try to do 
something behind its back, like stopping a VM or deleting it. We need to 
do it via "haizea-cancel-lease".

So, what happens (or what is supposed to happen) if:

1. I decide to stop or suspend or pause a VM via ONE: "onevm stop 123"?
2. Shutdown a VM via its own interface (ssh to Linux and issue "shutdown 
now" or RDP to Windows and click on "Shutdown")

What I have experienced so far, is that if I stop a VM via "onevm stop 
<ID>", then I cannot start it anymore. It stays in "pending" state 
forever. Is it waiting for the scheduler to "launch" it? Because if it 
is, Haizea will not do it, IMO - it has already launched that machine 
once, why launch it again?

On the other hand a shutdown would cause Haizea to become unstable, 
sometimes crashing the daemon, sometimes just loosing RPC. In the later 
case, I would see the daemon running and logs are filled, but 
"haizea-list-leases" would respond that it cannot communicate with the 
daemon.

So, how do theses VM life-cycle transitions fit into Haizea's view of 
the world?

I know that OpenNebula will list a VM in state "unknown" if it is 
manually shutdown. Which is dumb, since VMware is correctly showing the 
machine to be powered off.

If we want to be able to create, start and stop VMs as we see fit, how 
should we go about it, if Haizea is in the picture?

With ONE default scheduler, we successfully employed ONE XML-RPC. With 
Haizea, what should we use? System.exec("haizea-cancel-lease")? OK, we 
could use Jython and Haizea Python files, I guess. But still, do we have 
sufficient functionality now to execute these actions:

1. create a VM from template (that is OK, since we use it via ONE interface)
2. stop VM
3. resume VM
4. delete VM (should this be done via "haizea-cancel-lease")

Regards,
Nix.