On Fri, Sep 25, 2015 at 10:04 AM, Jan Schermer <jan@xxxxxxxxxxx> wrote: > I get that, even though I think it should be handled more gracefuly. > But is it expected to also lead to consistency issues like this? I don't think it's expected, but obviously we never reproduced it in the lab. Given that dumpling is EOL and the huge number of changes in all this code since then, I don't think you should expect anybody to figure out why it broke, though. :/ > > I think this is exactly what we're hitting right now > http://tracker.ceph.com/issues/6101 except I have no idea why it also > happens on a freshly backfilled OSD... Yeah, probably the backfill went through a different path that didn't fail but copied the bad data over. You could try using the patches on that ticket to get things up and running again... -Greg _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |