Hi James, the best practice is to set up 1 OSD daemon per physical disk drive. In your case, each OSD node would hence be 4 OSD daemons using one physical drive per daemon, and deploying a minimum of 3 servers so each object copy resides on a separate physical server. JC On Sep 25, 2014, at 20:42, James Pan <devjmp at yahoo.com> wrote: > Hi, > > I have several servers and each server has 4 disks. > Now I am going to setup Ceph on these servers and use all the 4 disks but it seems one OSD instance can be configured with one backend storage. > > So there seems two options to me: > > 1. Make the 4 disks into a raid0 then setup OSD to use this raid0 but obviously this is not good because one disk failure will ruin the entire storage. > 2. Build FS on each disk and start 4 OSD instances on the server. > > Both options are not good. So I am wondering what's the best practice of setting up multiple didks on one OSD for Ceph. > > > Thanks! > Best Regards, > > > > James Jiaming Pan > _______________________________________________ > ceph-users mailing list > ceph-users at lists.ceph.com > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |