When the compute host recovers from a failure the compute manager
destroys instances that were evacuated from the host while it was down.
However these code paths only consider evacuated instances that are
still reported by the hypervisor. This means that if the compute
host is recovered in a way that the hypervisor lost the definition
of the instances (for example the compute host was redeployed) then
the allocation of these instances will not be deleted.
This patch makes sure that the instance allocation is cleaned up
even if the driver doesn't return that instance as exists on the
hypervisor.
Note that the functional test coverage will be added on top as it needs
some refactoring that would make the bugfix non backportable.
Conflicts: nova/compute/manager.py
Conflict is due to not having I8ec3a3a697e55941ee447d0b52d29785717e4bf0
in Rocky. Also changes needed to be made in test_compute_mgr.py due to
I2af45a9540e7ccd60ace80d9fcadc79972da7df7 is missing form rocky.
Change-Id: I4bc81b482380c5778781659c4d167a712316dab4
Closes-Bug: #1724172
(cherry picked from commit 9cacaad14e8c18e99e85d9dc04308fee91303f8f)
(cherry picked from commit b8f2cd689f0a747778080ba4b6e148e71eb53085)
Reviewed: https:/ /review. opendev. org/687550 /git.openstack. org/cgit/ openstack/ nova/commit/ ?id=09de94e39bb cfc7f8130638e73 a8248e49cb6ab7
Committed: https:/
Submitter: Zuul
Branch: stable/rocky
commit 09de94e39bbcfc7 f8130638e73a824 8e49cb6ab7
Author: Balazs Gibizer <email address hidden>
Date: Tue Oct 17 15:06:59 2017 +0200
cleanup evacuated instances not on hypervisor
When the compute host recovers from a failure the compute manager
destroys instances that were evacuated from the host while it was down.
However these code paths only consider evacuated instances that are
still reported by the hypervisor. This means that if the compute
host is recovered in a way that the hypervisor lost the definition
of the instances (for example the compute host was redeployed) then
the allocation of these instances will not be deleted.
This patch makes sure that the instance allocation is cleaned up
even if the driver doesn't return that instance as exists on the
hypervisor.
Note that the functional test coverage will be added on top as it needs
some refactoring that would make the bugfix non backportable.
Conflicts:
nova/ compute/ manager. py 41ee447d0b52d29 785717e4bf0 e7ccd60ace80d9f cadc79972da7df7 is missing form rocky.
Conflict is due to not having I8ec3a3a697e559
in Rocky. Also changes needed to be made in test_compute_mgr.py due to
I2af45a9540
Change-Id: I4bc81b482380c5 778781659c4d167 a712316dab4 99e85d9dc04308f ee91303f8f) 778080ba4b6e148 e71eb53085)
Closes-Bug: #1724172
(cherry picked from commit 9cacaad14e8c18e
(cherry picked from commit b8f2cd689f0a747