On 01/14/2014 09:44 AM, Dietmar Maurer wrote: >>> When using a pool size of 3, I get the following behavior when one OSD >>> fails: >>> * the affected PGs get marked active+degraded >>> >>> * there is no data movement/backfill >> >> Works as designed, if you have the default crush map in place (all replicas must >> be on DIFFERENT hosts). You need to tweak your crush map in this case, but be >> aware that this can have serious effects (think of all your data residing on 3 disks >> on a single host). > > The old behavior was that data is automatically distributed to remaining 3 disks. > So the question is why this is different when we use 'ceph osd crush tunables optimal'? Sorry, it seems as if I had misread your question: Only a single OSD fails, not the whole server? Then there should definitively be a backfilling in place. Check if you have set the 'osd noout' flag in your cluster. -- http://www.wogri.com _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |