Reboot should not change state to active until after the reboot completes

[Phil Day]

reboot_instance() in the compute manager currently changes the vm_state to active before it does anything else.

If an instance was stopped (for example) and the reboot fails the vm_state is left as active

It should leave the vm_state alone until the reboot has successfully completed, and then change to Active if it is not Rescued.

It should also set vm_state to error if the reboot fails.

See also this bug as to why going to Error is acceptable:
https://bugs.launchpad.net/nova/+bug/1183946

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/32214

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/32214
Committed: http://github.com/openstack/nova/commit/bd9269278a532468e81bc752f89f05d74d9ee8e5
Submitter: Jenkins
Branch: master

commit bd9269278a532468e81bc752f89f05d74d9ee8e5
Author: Phil Day <email address hidden>
Date: Sat Jun 8 00:30:18 2013 +0100

    Fix bad vm_state change in reboot_instance()

    Currently reboot_instance() in the compute manager
    changes the vm_state to active before it does anything
    else. This means that if there is an error in the reboot
    operation then the vm_state is left as active although
    the instance may not be running.

    This change leaves the vm_state alone until the reboot
    has completed

    If the reboot fails but the power_state shows that it
    is running then we log the fault but leave it in an
    active or rescued state

    If the reboot fails and the power_state shows that it
    is not running then we set vm_state to error

    See also https://bugs.launchpad.net/nova/+bug/1183946
    which make recovery from an Error state possible.

    Fixes bug: 1188343

    Change-Id: Idabd358d3bcaf63e21ff59188f805143dbaf9ada

[Swami Reddy (swamireddy)]

Could not see the patch at https://review.openstack.org/c/131374/ . How can I see this patch?