Hi Sage, When the OSD is prepared, we add a --device-class option to ceph-disk that creates a device_class file and activate can use it use when creating the the OSD[1]. The ceph osd create command is given an additional optional argument (device class). If the device class argument is present, ceph osd create[2] creates a device entry in the crushmap (it currently does nothing with the crushmap) so that the corresponding device class can be preserved. This happens before update_crush_location[3] is called and since the device is not included in any crush tree it won't be used. I'm not happy about adding an argument to osd create. We could instead have a new dedicated command create-with-device-class but I like that even less. We could also add a "ceph osd crush create-device" and have ceph-disk call it right after "ceph osd create", followed by "ceph osd crush set-device-class". That would be my second favorite: the only drawback is having a multi-step device creation during activation but all steps are idempotents that should not be a problem. What do you think ? Cheers [1] https://github.com/ceph/ceph/blob/master/src/ceph-disk/ceph_disk/main.py#L1055 [2] https://github.com/ceph/ceph/blob/master/src/mon/OSDMonitor.cc#L7604 [3] https://github.com/ceph/ceph/blob/master/src/osd/OSD.cc#L3024 -- Loïc Dachary, Artisan Logiciel Libre -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
