On Tue, Jun 5, 2018 at 4:46 PM, shrey chauhan <shrey.chauhan@xxxxxxxxxxxxx> wrote: > I am consistently getting whiteout mismatches due to which pgs are going in > inconsistent state, and I am not able to figure out why is this happening? > though as it was explained before that whiteouts dont exist and its nothing, > its still painful to see my pgs in inconsistent state....can any one help me > with this, whats the work around. It's not that the whiteouts don't exist, they represent objects that don't exist in the base pool so when an IO request comes in for an object that has a whiteout the cache tier pool does not have to check the base pool because it already knows the object does not exist and can respond quicker. > > when do these whiteouts occur? When an IO request is made on an object that does not exist in the base pool. > > cluster [ERR] 9.3d deep-scrub stat mismatch, got 803/803 objects, 0/0 > clones, 791/791 dirty, 42/42 omap, 0/0 pinned, 12/12 hit_set_archive, > 147/145 whiteouts, 13468918/13468918 bytes, 4649/4649 hit_set_archive > bytes.. This is likely an accounting error. The first step in getting this rectified is to open up a tracker so it can be prioritised and analysed further. -- Cheers, Brad _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |