Instances stuck in Image_snapshot/Queued should be cleaned up

[Mandar Vaze]

Scenario:
If various systems like RabbitMQ, mysql DB server ot nova-compute go down during snapshot, the instance is stuck either in queued or Image_snapshot state. (Depending on the timing of when a specific component went down)

Expected Response:
Instance should be eventually brought back to ACTIVE state.
If there are snapshot entries in glance DB and/or on disk, they should be cleaned up.

Actual Response:
Instance remains stuck in Image_snapshot state.
This is problematic because once it is stuck in this state, no snapshot is allowed on this instance till it returns in either ACTIVE or SHUTOFF state.

    in nova/compute/api.py
    @check_instance_state(vm_state=[vm_states.ACTIVE, vm_states.SHUTOFF])
    def snapshot(self, context, instance, name, extra_properties=None):

Notes :
This was reproduced forcefully for testing purposes by putting breakpoint at appropriate place(s) and then shutting down rabbitmq or mysql servers from other terminal window.

Branch: master

[Mandar Vaze (mandarvaze)]

linked to blueprint https://blueprints.launchpad.net/nova/+spec/compute-instance-cleanup-service

[Sean Dague (sdague)]

Confirmed as it is part of a blueprint

[Joe Gordon (jogo)]

Is this still valid? The blueprint was superseded.