On Thu, Jul 4, 2013 at 7:09 PM, huangjun <hjwsm1989@xxxxxxxxx> wrote: > hi,all > there are some questions about ceph. > 1) can i get the osd list that hold objects consist of a file by > command-line? If using CephFS you can use the cephfs tool to map offsets to locations. Is that what you mean? > 2) does ceph support FC or how can i get FC supported in ceph? like we use > iscsi target tool to export rbd block device ,so windows can use it as a > disk. Hmm, don't think so yet. Does that concept even make sense for fibre channel connections? > 3) the question i thought many times, the crush hierarchy is > stack,host,device,and the replication level is 2, but if two disks failed in > different stack in 5 mins, how can i cut down the risk of data lose to > lowest? I thought about disaster backup by building a remote datacenter, > but if another two disks also failed in that datacenter? The possibility of data loss due to simultaneous disk failures is pretty much a constant (though fairly unlikely!). If two disks doesn't provide the redundancy you need, you can use three. Separating the failure domains as much as possible of course helps reduce the odds more. > 4) "df -h" on client shows the data stored(include the replicated data) in > cluster, not data really used by user. Yep. Keeping in mind that due to using different replication settings etc we can't map the total space to a user-visible space, what would you rather see here? -Greg Software Engineer #42 @ http://inktank.com | http://ceph.com _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |