Bug #34628
closedceph health detail truncated output
0%
Description
Hi,
Ceph 13.2.1 here.
It seems to me the command "ceph health detail" is outputing a "limited number of details".
This is what I see :
[root@ceph0 ~]# ceph health detail |head -n 3 HEALTH_ERR 2956 scrub errors; Possible data damage: 510 pgs inconsistent OSD_SCRUB_ERRORS 2956 scrub errors PG_DAMAGED Possible data damage: 510 pgs inconsistent [root@ceph0 ~]# ceph health detail |tail -n +4|wc -l 51 [root@ceph0 ~]# ceph health detail |grep "active+clean+inconsistent"|wc -l 51
I.E : only 51 inconsistent PGs displayed, but ceph reports 510 damaged.
JSON output does not show more :
[root@ceph0 ~]# ceph health detail --format json-pretty|jq -r '.checks.PG_DAMAGED.detail[].message'|wc -l 51 [root@ceph0 ~]# ceph health detail --format json-pretty|jq -r '.checks.PG_DAMAGED.detail[]'|head { "message": "pg 11.3a1 is active+clean+inconsistent, acting [189,24,113]" } (...)
Expected output : all the damaged PGs ?
I know I have that damn number of 510 inconsistent PGs: I've run a deep scrub on all of the PGs pool since that number did not stop growing after the 12.2.6 bug hit and after I upgraded to 12.2.7 (and now mimic), and since then I've been repairing the PGs by overwritting the "damaged" objects to clear the errors (and I could get down to 0 bad PGs until more scrubs ran)
Regards
Updated by John Spray over 5 years ago
- Status changed from New to Rejected
This is not a bug -- we limit the detail fields to avoid generated unbounded-size health output. The setting that controls how many PGs may be reported in the health output is mon_health_max_detail