Instance is of no use during "shelve-offloading" operation, when nova-conductor is stopped

Bug #1774363 reported by Swati Bansal
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
OpenStack Compute (nova)
New
Undecided
Unassigned

Bug Description

During shelve-offloading operation if nova-conductor is stopped, the instance stuck forever in "SHELVED" state and task state of instance stuck in "SHELVING_OFFLOADING" state.
After restarting the nova-conductor service, the state of instance does not change to "SHELVE_OFFLOADED" from "SHELVED" and the task state remains in "SHELVING_OFFLOADING" state. In this particular state, the instance is of no use because an instance cannot shelve, unshelve and shelve-offload.

Source code should be modified to reset the state of instance.

tags: added: shelve
tags: added: conductor
Revision history for this message
Matt Riedemann (mriedem) wrote :

Everything goes through nova-conductor for database updates so if that's down, you have problems. You can reset the state of the instance using the reset state API (as admin):

https://docs.openstack.org/python-openstackclient/latest/cli/command-objects/server.html#server-set

nova server set --state error

Although you can't unshelve from that if the instance is really offloaded and not on a host. If the shelve offload image snapshot was successful, then you basically need to delete the old server and create a new one from the shelved offloaded image snapshot.

Changed in nova:
status: New → Invalid
Revision history for this message
Matt Riedemann (mriedem) wrote :

What it sounds like you're asking for is the ability to rebuild a shelved offloaded instance in error state, which would actually trigger something like an evacuate operation to build the instance on a new host (like unshelve). That's not supported today, and to add that support would be an API change to rebuild and require a spec:

https://docs.openstack.org/nova/latest/contributor/blueprints.html

Revision history for this message
Ian Kumlien (pomac) wrote :

This is still happening btw, and there is no good way out.

Risking the VM is not always an option.

Ian Kumlien (pomac)
Changed in nova:
status: Invalid → New
Revision history for this message
Ian Kumlien (pomac) wrote :

Do, in this particular case, it was stuck in "shelved" - however the disk had been shelved properly, so *boom* another vm bites the dust.

Not having a way out of shelve/shelved_offloaded if openstack deems something odd is broken.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.