The exclusive-lock feature does, by default, automatically transition the lock between clients that are attempting to use the image. Only one client will be able to issue writes to the image at a time. If you ran "dd" against both mappings concurrently, I'd expect you'd see a vastly decreased throughput due to this lock exchange. There is a pending change to optionally disable this automatic lock transition that it is expected be included in a future kernel release. On Tue, Mar 21, 2017 at 6:54 AM, Mikaël Cluseau <mcluseau@xxxxxx> wrote: > Hi, > > There's something I don't understand about the exclusive-lock feature. > > I created an image: > > $ ssh host-3 > Container Linux by CoreOS stable (1298.6.0) > Update Strategy: No Reboots > host-3 ~ # uname -a > Linux host-3 4.9.9-coreos-r1 #1 SMP Tue Mar 14 21:09:42 UTC 2017 x86_64 Intel(R) Xeon(R) CPU E5-2620 v3 @ 2.40GHz GenuineIntel GNU/Linux > host-3 ~ # rbd create test-xlock --size=1024 --image-feature exclusive-lock > host-3 ~ # rbd feature enable test-xlock exclusive-lock > rbd: failed to update image features: (22) Invalid argument2017-03-21 10:16:50.911598 7f6975ff0100 -1 librbd: one or more requested features are already enabled > > I mapped it > > host-3 ~ # rbd map --options lock_on_read test-xlock > /dev/rbd0 > > I also could map it from another host (disappointment started here): > > host-2 ~ # rbd map --options lock_on_read test-xlock > /dev/rbd9 > > I can read from both host: > > host-2 ~ # dd if=/dev/rbd9 of=/dev/null > 2097152+0 records in > 2097152+0 records out > 1073741824 bytes (1.1 GB, 1.0 GiB) copied, 4.61844 s, 232 MB/s > > host-3 ~ # dd if=/dev/rbd0 of=/dev/null > 2097152+0 records in > 2097152+0 records out > 1073741824 bytes (1.1 GB, 1.0 GiB) copied, 3.33527 s, 322 MB/s > > And also write: > > host-3 ~ # dd if=/dev/urandom of=/dev/rbd0 bs=1M count=1 > 1+0 records in > 1+0 records out > 1048576 bytes (1.0 MB, 1.0 MiB) copied, 0.0153388 s, 68.4 MB/s > > host-2 ~ # dd if=/dev/urandom of=/dev/rbd9 bs=1M count=1 > 1+0 records in > 1+0 records out > 1048576 bytes (1.0 MB, 1.0 MiB) copied, 0.0312396 s, 33.6 MB/s > > Isn't exclusive-lock supposed to forbid at least concurrent writes? > > _______________________________________________ > 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
 |