[ceph-users] OSD_ORPHAN issues after jewel->luminous upgrade

Graham Allan gta at umn.edu
Thu Dec 7 10:36:44 PST 2017

Just updated a fairly long-lived (originally firefly) cluster from jewel 
to luminous 12.2.2.

One of the issues I see is a new health warning:

OSD_ORPHAN 3 osds exist in the crush map but not in the osdmap
     osd.2 exists in crush map but not in osdmap
     osd.14 exists in crush map but not in osdmap
     osd.19 exists in crush map but not in osdmap

Seemed reasonable enough, these low-numbered OSDs were on 
long-decommissioned hardware. I thought I had removed them completely 
though, and it seems I had:

> # ceph osd crush ls osd.2
> Error ENOENT: node 'osd.2' does not exist
> # ceph osd crush remove osd.2
> device 'osd.2' does not appear in the crush map

so I wonder where it's getting this warning from, and if it's erroneous, 
how can I clear it?

Graham Allan
Minnesota Supercomputing Institute - gta at umn.edu

More information about the ceph-users mailing list