On Thu, Oct 19, 2017 at 12:59 AM, zhaomingyue <zhao.mingyue@xxxxxxx> wrote: > Hi: > > when I analyzed the performance of ceph, I found that rebuild_aligned was > time-consuming, and the analysis found that rebuild operations were > performed every time. > > > > Source code: > > FileStore::queue_transactions > > –> journal->prepare_entry(o->tls, &tbl); > > -> data_align = ((*p).get_data_alignment() - bl.length()) & ~CEPH_PAGE_MASK; > > -> ret = ebl.rebuild_aligned(CEPH_DIRECTIO_ALIGNMENT); > > > > Log: > > 2017-10-17 19:49:29.706246 7fb472bfe700 10 journal len 4196131 -> 4202496 > (head 40 pre_pad 4046 bl 4196131 post_pad 2239 tail 40) (bl alignment 4086) > > > > question: > > I find “alignment =4086”, I think It maybe a bug > > I think it should be 4096 > > because CEPH_DIRECTIO_ALIGNMENT is 4096 What led you to this, and what version of the code are you running/examining? I don't see any instance of "4086" in the codebase, so there's not a typo, CEPH_DIRECTIO_ALIGNMENT is set to 4096, and our invocations of rebuild_aligned in master do not look like the code snippets you've got there that I can see. -Greg _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |