interrupted update on DB/image store may lead to resource leakage

During image registration/deletion, if the update of DB state or image store is interrupted (e.g., by fault), then it may lead to resource leakage.

Below are two examples:

1. glance registration example
1.1 issue command "glance image-create" with "copy-from=http" option
1.2 image is copied to the glance host
1.3 image state in glance DB is updated

If a fault occurs between 1.2 and 1.3 (e.g., glance service crashes and then restarts), then the image may exist on glance host but is not correctly recorded in the DB. Subsequent image deletion cannot remove such an image from the glance host (in particular because the image location information in DB may be incorrect in such cases).

2. glance deletion example
2.1 issue command "glance image-delete"
2.2 image state in glance DB is updated
2.3 image is removed from the glance host

If a fault occurs between 2.2 and 2.3, then the image record may be removed from DB but the image itself remains on the glance host.