c2b7b78d4f
When updating legacy images, currenly we use the user's context and elevate priviledges. However, we do not require admin priviledges for the cinder API calls. This patch removes the special case where we elevate priviledges as it wasn't doing anything rather avoiding us to use right credentials and failing to fetch volume in the right location because of wrong credentials. The correct credentials are either the service ones set in glance-api.conf file or the user context credentials, using which the Image-Volume was created. NOTE: When using cinder as glance backend and we want to perform optimized volume upload to image, one thing we should make sure is either using the context or the cinder credentials set in glance-api.conf file, it should match the following details on the cinder side (if we are using internal context to create clone of image-volumes): cinder_store_user_name = context.user_id/cinder_internal_tenant_user_id cinder_store_project_name = context.project_id/cinder_internal_tenant_project_id The cinder_internal_tenant_user_id and cinder_internal_tenant_project_id are set in the [DEFAULT] section of cinder.conf. This issue was first discovered when testing the new location APIs[1] where tempest creates a volume with cinder's internal context and glance uses wrong (user context) credentials to access it and failing with 404 not found. [1] https://review.opendev.org/c/openstack/cinder/+/909847 Closes-Bug: #2056179 Change-Id: I4f27a9800f239da8dbf29f4c028678df1f867664 |
||
|---|---|---|
| .. | ||
| cinder | ||
| swift | ||
| __init__.py | ||
| filesystem.py | ||
| http.py | ||
| rbd.py | ||
| s3.py | ||
| vmware_datastore.py | ||