[nova] instance not recoverable when its image backend got deleted

Sean Mooney smooney at redhat.com
Fri Feb 3 22:54:08 UTC 2023


On Fri, 2023-02-03 at 15:16 -0600, hai wu wrote:
> For an Openstack VM instance, if its backing image got deleted
> already, and this VM instance was already hard rebooted, and went into
> `Error` state due to `Image not found` error.
a hard reboot does not try to download the image or even interact with glance
if the instance went into error that likely means something esle happend that
caused that.

in general deleteing an image form glance will have no impact on a running instnace and hard
reboots or live migration will continue to work.

evacuate or rebuild/rescue (without sepcifying a new image)  will fail unless the image was
boot form volume.
the fact the image is deleted in glance is proably not the root cause of the issue but it will limit what
you can do to recover.

if the iamge is in use by a vm on the host and you are not using images_type=flat  (i.e. you are using images_type=qcow or raw) the image shoudl still
be in the compute nodes image cache. as an  admin you could undelete the iamge in database and reupload it.

if the vm has a disk that  refecenes a backing file and that file nolonger exist on the compute host or glance there is not way to fully recover the
vm you might be able to scp the file somewere and manualy extract teh vms data form the delta disk but there is nothign you can do form the
openstack api to fix this.
> 
> Is it possible to recover this VM stucking in this bad state? I tried
> to create a vm snapshot, and then try to rebuild this VM using the
> newly created vm snapshot, and it kept failing.
> 
> Regards,
> Hai
> 




More information about the openstack-discuss mailing list