On Thu, Feb 15, 2018 at 6:40 AM, Vladimir Prokofev <v@xxxxxxxxxxx> wrote: > CEPH 12.2.2 > > So I have a snapshot of an image: > > $ rbd du ssd-volumes/volume-4fed3f31-0802-4850-91bc-17e6da05697d > NAME > PROVISIONED USED > volume-4fed3f31-0802-4850-91bc-17e6da05697d@snapshot-d2145e21-99a7-4e2e-9138-ab3e975f8113 > 20480M 8088M > volume-4fed3f31-0802-4850-91bc-17e6da05697d > 20480M 1152M > <TOTAL> > 20480M 9240M > > I create a clone, and then flatten it. > > Here's a clone before flatten: > $ rbd info ssd-volumes/volume-0578b38a-8db2-481f-ad43-65d21b09c89b > rbd image 'volume-0578b38a-8db2-481f-ad43-65d21b09c89b': > size 20480 MB in 5120 objects > order 22 (4096 kB objects) > block_name_prefix: rbd_data.33717c18c5a7e1 > format: 2 > features: layering, exclusive-lock, object-map, fast-diff, > deep-flatten > flags: > parent: > ssd-volumes/volume-4fed3f31-0802-4850-91bc-17e6da05697d@snapshot-d2145e21-99a7-4e2e-9138-ab3e975f8113 > overlap: 20480 MB > > $ rbd du ssd-volumes/volume-0578b38a-8db2-481f-ad43-65d21b09c89b > NAME PROVISIONED USED > volume-0578b38a-8db2-481f-ad43-65d21b09c89b 20480M 0 > > After flatten: > $ rbd info ssd-volumes/volume-0578b38a-8db2-481f-ad43-65d21b09c89b > rbd image 'volume-0578b38a-8db2-481f-ad43-65d21b09c89b': > size 20480 MB in 5120 objects > order 22 (4096 kB objects) > block_name_prefix: rbd_data.33717c18c5a7e1 > format: 2 > features: layering, exclusive-lock, object-map, fast-diff, > deep-flatten > flags: > > $ rbd du ssd-volumes/volume-0578b38a-8db2-481f-ad43-65d21b09c89b > NAME PROVISIONED USED > volume-0578b38a-8db2-481f-ad43-65d21b09c89b 20480M 20480M > > Clone is now using 100% provisioned space. > I was under impression that flattened clones should be sparse. Am I wrong or > missing something? Since the Luminous release, the flatten operation should avoid de-sparsifying the cloned image if the backing parent object is fully zeroed. However, if at least one byte is non-zero in the backing parent image, it will result in a fully allocated clone block since we do not yet support a sparse "copy-up" write operation from parent to clone (although it's something we would like to address in the future along with a tool to re-sparsify images). > Don't think this is relevant, but snapshot and clone creation are managed by > cinder in Openstack Newton release. > > _______________________________________________ > ceph-users mailing list > ceph-users@xxxxxxxxxxxxxx > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > -- Jason _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |