Hi, just make sure you modify your CRUSH map so that each copy of the objects are just dispatched on different OSDs rather than on different hosts. Follow these steps: ceph osd getcrushmap -o /tmp/cm crushtool -i /tmp/cm -o /tmp/cm.txt Edit the /tmp/cm.txt file. Locate the crush rule ID 0 at the end of the text file and replace "chooseleaf firstn 0 type host" with "chooseleaf firstn 0 type osd" crushtool -c /tmp/cm.txt -o /tmp/cm.new ceph osd setcrushmap -i /tmp/cm.new And this should do the trick. Cheers JC > On 18 Mar 2015, at 09:57, Khalid Ahsein <kahsein@xxxxxxxxx> wrote: > > Hello everybody, > > I want to build a new architecture with Ceph for storage backend. > For the moment I’ve got only one server with this specs : > > 1 RAID-1 SSD : OS + OSD journals > 12x 4To : OSD daemons. > > I never reached the « clean state » on my cluster and I’m always in HEALTH_WARN mode like this : > health HEALTH_WARN 25 pgs degraded; 24 pgs incomplete; 24 pgs stuck inactive; 64 pgs stuck unclean; 25 pgs undersized > > I tried to add 3 —> 12 OSD but it’s always the same problem. > > What is the right configuration to have a valid cluster please ? > > # cat ceph.conf > [global] > fsid = 588595a0-3570-44bb-af77-3c0eaa28fbdb > mon_initial_members = drt-marco > mon_host = 172.16.21.4 > auth_cluster_required = cephx > auth_service_required = cephx > auth_client_required = cephx > filestore_xattr_use_omap = true > public network = 172.16.21.0/24 > > [osd] > osd journal size = 10000 > osd crush chooseleaf type = 0 > osd pool default size = 1 > > NB : I use ceph-deploy for debian wheezy to deploy the services. > > Thank you so much for your help ! > k. > _______________________________________________ > 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
 |