The journal should be a raw partition and should not have any filesystem on it. Inside your /var/lib/ceph/osd/ceph-# you should make symlink to the journal partition that you are going to use for that osd. On Thu, Jun 18, 2015 at 2:36 AM, Shane Gibson <Shane_Gibson@xxxxxxxxxxxx> wrote: > All - I am building my first ceph cluster, and doing it "the hard way", > manually without the aid of "ceph-deploy". I have successfully built the > mon cluster and am now adding OSDs. > > My main question: > How do I prepare the "Journal" prior to the prepare/activate stages of the > OSD creation? > > > More details: > Basically - all of the documentation seems to "assume" the journal is > "prepared". Do I simply create a single raw partition on a physical > device and the "ceph-disk prepare..." and "ceph-disk activate..." steps > will "take care" of everything for the journal ... presumably based on the > "ceph-disk prepare ... --type" filesystem setting? Or do I need to > actually format it as a filesystem prior to giving it over to the Ceph OSD > ??? > > The architecture I'm thinking of is as follows - based on the hardware I > have for OSDs (currenly 9 servers each with): > > RAID 0 mirror for OS hard drives (2 disks) > data disk for journal placement for 5 physical disks (4TB) > data disk for journal placement for 5 physical disks (4TB) > 10 data disks as OSDs (one OSD per disk) (4TB each) > > Essentially - there are "12 data disks" in the node (all 4 TB 7200 rpm > spinning disks). Splitting the Journal across two of them gives me a > failure domain of "5 data disks + 1 journal disk" in a single physical > server for crush map purposes ... It also vaguely helps spread the I/O > workload for the journaling activity across 2 physical disks in a chassis > instead of a one (since the journal disk is "pretty darn slow"). > > In this configuration I'd create 5 separate partitions on Journal Disk A > and 5 on Journal Disk B ... but do they need to be formatted and mounted? > > Yes, we know as we go to more "real production" workloads, we'll want/need > to change this for performance reasons - eg the Journal on SSDs ... > > Any pointers on where I missed this info in the documentation would be > helpful too ... I've been all over the ceph.com/docs/ site and haven't > found it yet... > > Thanks, > ~~shane > > _______________________________________________ > 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
 |