I have a cluster running 15.2.1, was originally running 14.x, the cluster is running the balance module in upmap mode (I have tried crush-compat in the past) Most OSD's are around the same & used give or take 0.x, however there is one OSD that is down a good few % and a few that are above average by 1 or 2 %, I have been trying to get the balance to fix this. I have tried running a manual osdmaptool command on an export of my map, but it lists no fixed however does display the underfall OSD in it's output (overfull 3,4,5,6,7,8,9,10,11,12,13,14,15,18,19,20 underfull [36]) The debug output is just lots of: 2020-05-05T06:15:39.172+0000 7f3dfb0c3c40 10 trying 2.55 2020-05-05T06:15:39.172+0000 7f3dfb0c3c40 10 2.55 [12,3,7,6,33,34,30,35,21,18] -> [12,3,7,6,33,34,30,35,21,16] 2020-05-05T06:15:39.172+0000 7f3dfb0c3c40 10 will try adding new remapping pair 18 -> 16 for 2.55 NOT selected osd 2020-05-05T06:15:39.172+0000 7f3dfb0c3c40 10 stddev 528.667 -> 528.667 2020-05-05T06:15:39.172+0000 7f3dfb0c3c40 10 Overfull search osd.7 target 170.667 deviation 9.33327 Is there anything I can to try and balance the overfull onto the underful OSDs to balance out the last bit. _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |