[one-users] How to bring back a failed VM?

[Ruben S. Montero]

Hi,

We have here a couple of different issues:

* Delete images of VMs in FAILED state. The rationale behind this is
not to leave unneeded images in the worker-nodes that would eventually
fill the node disks or the $ONE_LOCATION/var directory. However this
is not always the desired behavior (e.g. checkpoint files), specially
now that we have a delete action.

* Manually restoring VMs should only occur in the UNKNOWN state. In
this case OpenNebula still polls the VM and would automatically
recover it. In fact, one of the reasons for UNKNOWN is to let the
admin manually recover a VM under unexpected circumstances, this
should only happen for VMs that were running, and not in a failed
state.

* Re-use of a VM, in failed state. Some times it is useful to re-use a
VM that is in FAILED state. Specially when you are debugging the
infrastructure. The option here is to re-submit the original template.
However this is not useful on some use cases that needs to preserve
the ID of the VM or the network leases. In this case we could probably
transition the VM from FAILED to PENDING.

Some of these changes affect core modules of the OpenNebula daemon and
cannot be done at this stage of the release process. So, there will
not be any additional modifications to the life-cycle of a VM in 2.0.

We also need to better understand the modifications we require, this
can happen just after the 2.0.

Cheers

Ruben



On Tue, Sep 7, 2010 at 10:43 PM, Tres Wong-Godfrey <tres at blas.phemo.us> wrote:
>
> On Sep 7, 2010, at 1:35 PM, Shi Jin wrote:
>
>> Right. I actually commented out the delete action since some failure shouldn't cause a deletion.
>
> Yeah, I actually did the same. I'm hoping to see a "PERSISTENT_VM" lifecycle or something that could be configured from the one conf file.
>
>> I guess there is no way in 1.4 then.
>
> A workaround is to build a new host, then rename disks from the old host. Get's things plugged back in.
>
>> I hope in 2.0 this is handled more gracefully.
>
> Me too. I completely understand the design -- a canonical cloud host is not supposed to be persistent. But it really confines OpenNebula to a very limited role.
>
> Regards,
> Tres
>
>
>
> _______________________________________________
> Users mailing list
> Users at lists.opennebula.org
> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
>



-- 
Dr. Ruben Santiago Montero
Associate Professor (Profesor Titular), Complutense University of Madrid

URL: http://dsa-research.org/doku.php?id=people:ruben
Weblog: http://blog.dsa-research.org/?author=7