About incurred case, I propose not to modify glance code, but create a script, that shows all duplicate locations in the deployment and warns the operator if there is something wrong.
It's better because, because once my code's merged there is no possibility to set duplicate locations, so Fei Long patch seems redundant. Also, correct me if I'm wrong, there are two possibilities to set links to one image data in ceph clusters: 'rbd://<image_id>' and 'rbd://<fsid>/pool/image/<image_id>', so 'image_locations_with_url' may return wrong result and we have to handle it somehow.
Thanks for your comments, folks!
I updated the patch, so please check it asap.
About incurred case, I propose not to modify glance code, but create a script, that shows all duplicate locations in the deployment and warns the operator if there is something wrong. <fsid>/ pool/image/ <image_ id>', so 'image_ locations_ with_url' may return wrong result and we have to handle it somehow.
It's better because, because once my code's merged there is no possibility to set duplicate locations, so Fei Long patch seems redundant. Also, correct me if I'm wrong, there are two possibilities to set links to one image data in ceph clusters: 'rbd://<image_id>' and 'rbd://