On Wed, May 8, 2019 at 7:26 AM <jesper@xxxxxxxx> wrote: > > Hi. > > I'm fishing a bit here. > > What we see is that when we have new VM/RBD/SSD-backed images the > time before they are "fully written" first time - can be lousy > performance. Sort of like they are thin-provisioned and the subsequent > growing of the images in Ceph deliveres a performance hit. Do you have object-map enabled? On a very fast flash-based Ceph cluster, the object-map becomes a bottleneck on empty RBD images since the OSDs are only capable of performing ~2-3K object map updates / second. Since the object-map is only updated when a backing object is first written, that could account for initial performance hit. However, once the object-map is updated, it is no longer in the IO path so you can achieve 10s of thousands of writes per second. > Does anyone else have someting similar in their setup - how do you deal > with it? > > KVM based virtualization, Ceph Luminous. > > Any suggestions/hints/welcome > > Jesper > > _______________________________________________ > 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
 |