Original image_ref is lost after a failed unshelve
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Compute (nova) |
Fix Released
|
Undecided
|
Pierre-Samuel LE STANG |
Bug Description
Description
===========
When an unshelve fails the original image_ref of an instance is lost and replaced by the image id of the shelved image.
Steps to reproduce
==================
1/ Shelve an instance
2/ Unshelve the instance and abort it (kill the process on compute side) to generate an error
3/ If you unshelve again successfully the instance image_ref is the id of the shelved image and is marked as not found because it was deleted at the end of the unshelve process
Expected result
===============
Expect the instance image_ref to be the original image_ref
Actual result
=============
the instance image_ref is the one of the shelved image
Bug identified on Newton and Stein but seems still present on latest release
Changed in nova: | |
assignee: | nobody → Pierre-Samuel LE STANG (pslestang) |
I'm not sure that killing n-cpu is a valid usecase here, unshelve is a cast so you can't expect anything to cleanup outside of the n-cpu process you just killed.
Can you provide an example trace of the second unshelve request?