[Openstack-operators] no VM "really" up after complete hardware reboot

Andrew Glen-Young andrew.glen-young at canonical.com
Thu May 24 09:32:02 UTC 2012

On Thu 24-05-2012 10:57 ›, Christian Parpart wrote:
> On Thu, May 24, 2012 at 10:49 AM, Diego Parrilla Santamaría <
> diego.parrilla.santamaria at gmail.com> wrote:
> > You have two interesting parameters you can use to automatically launch
> > the images after reboot:
> >
> > resume_guests_state_on_host_boot
> > start_guests_on_host_boot
> >
> Yep, I first thought it might be something like this, too, however, the
> dashboard actually pretends that they're up.
> but nothing is running. The `nova list` output even says that they're
> running, too, but they aren't :)
> So I then tried to push the "reboot" button in the dashboard, and since
> then they're in infinite "restarting" state.

The running status may be a result of the 'reboot' definition in the
compute API.

The VM state is first set to active and then the VM is told to reboot.
If the VM fails to come up, the database will have the VM marked as
active even if it really is not.


    def reboot(self, context, instance, reboot_type):
        """Reboot the given instance."""
        state = {'SOFT': task_states.REBOOTING,
                 'HARD': task_states.REBOOTING_HARD}[reboot_type]
        self._cast_compute_message('reboot_instance', context, instance,
                params={'reboot_type': reboot_type})

To get the system into a good state, I have needed to terminate the
instances. The termination throws an exception on the compute nodes, but
the VMs are no longer marked as running.

> This somehow also crunched my nova.projects table (again). <--- veeeery odd.

Due to the above explanation, I do not believe this to be related [1][2].
It is, however worrying.

[1]: http://xkcd.com/552/
[2]: http://en.wikipedia.org/wiki/Correlation_does_not_imply_causation

Andrew Glen-Young

More information about the Openstack-operators mailing list