Hi! Because distribution is computed with CRUSH rules algoritmically. So, as with any other hash algorithms, the result will depend of the 'data' itself. In ceph, 'data' - is the object name. Imagine, that you have a simple plain hashtable with 17 buckets. Bucket index is computed by a simple 'modulo 17' algorithm. When you try to insert in the table values 17,34,51, etc, they will be stored in one bucket only, leaving all others empty. The same thing happens with ceph, except that the 'hash function' (CRUSH map) are heavily parametrized by osd tree topology, weights, etc. Also you can set number of placement groups per pool - pg num and pgp num. The higher their values, the more uniform will be distribution. But it is also rise the resources needed for monitors. Megov Igor CIO, Yuterra ________________________________________ От: ceph-users <ceph-users-bounces@xxxxxxxxxxxxxx> от имени Квапил, Андрей <kvaps@xxxxxxxxxxx> Отправлено: 7 сентября 2015 г. 14:44 Кому: ceph-users@xxxxxxxxxxxxxx Тема: Re: Ceph cache-pool overflow Hello. Somebody can answer a simple question? Why ceph, with the equal weights and sizes of OSD, writes to them not equally? - to some bit more, to other bit less... Thanks. _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |