Change I7b8622b178d5043ed1556d7bdceaf60f47e5ac80 started deleting
the associated resource provider when a compute service is deleted.
However, the delete_resource_provider cascade=True logic only looks
for instances on the given compute service host being deleted which
will miss (1) allocations remaining from evacuated servers and
(2) unconfirmed migrations.
Attempting to delete the resource provider results in an
ResourceProviderInUse error which delete_resource_provider ignores
for legacy reasons. This results in the compute service being
deleted but the resource provider being orphaned. What's more,
attempting to restart the now-deleted compute service will fail
because nova-compute will try to create a new resource provider
with a new uuid but with the same name (based on the hypervisor
hostname). That failure is actually reported in bug 1817833.
NOTE(mriedem): Note that in this backport a simple version of
assertFlavorMatchesUsage is added since the original version from
change If6aa37d9b6b48791e070799ab026c816fda4441c is not in Rocky.
Change-Id: I69f52f1282c8361c9cdf90a523f3612139cb8423
Related-Bug: #1829479
Related-Bug: #1817833
(cherry picked from commit 2629d65fbc)
(cherry picked from commit b18e42d20b)
6eda7409ff