维护ceph集群

1. set: 设置标志位

# ceph osd set <flag_name> # ceph osd set noout # ceph osd set nodown # ceph osd set norecover

2. unset: 清除标志位

# ceph osd unset <flag_name> # ceph osd unset noout # ceph osd unset nodown # ceph osd unset norecover

3. 标志位

noout: 该标志位将使 ceph 集群不会将任何 OSD 标记为 out(集群外)，无论其实际状态如何。这将会把所有的 OSD 保留在 ceph 集群中。

nodown: 该标志位将使得 ceph 集群不会将任何 OSD 标记为 down(服务已停止)，无论其实际状态如何。这将会使集群中的所有 OSD 保持 UP(服务运行中)状态，面不会是 DOWN 状态。

noup:

4. "Too many repaired reads on 1 OSDs" 告警处理

# ceph -s cluster: id: dfcdf8de-f388-4c84-adc2-ee721da8df84 health: HEALTH_WARN nodeep-scrub flag(s) set Too many repaired reads on 1 OSDs 3 pgs not deep-scrubbed in time services: mon: 1 daemons, quorum server (age 9w) mgr: server(active, since 9w) osd: 4 osds: 3 up (since 9w), 3 in (since 4M) flags nodeep-scrub data: pools: 3 pools, 320 pgs objects: 155.80k objects, 605 GiB usage: 1.2 TiB used, 3.4 TiB / 4.6 TiB avail pgs: 320 active+clean io: client: 0 B/s rd, 19 KiB/s wr, 1 op/s rd, 1 op/s wr

# ceph health detail HEALTH_WARN nodeep-scrub flag(s) set; Too many repaired reads on 1 OSDs; 3 pgs not deep-scrubbed in time OSDMAP_FLAGS nodeep-scrub flag(s) set OSD_TOO_MANY_REPAIRS Too many repaired reads on 1 OSDs osd.3 had 13 reads repaired PG_NOT_DEEP_SCRUBBED 3 pgs not deep-scrubbed in time pg 4.48 not deep-scrubbed since 2025-02-05 07:05:25.334392 pg 5.44 not deep-scrubbed since 2025-02-05 07:33:33.616573 pg 6.29 not deep-scrubbed since 2025-02-05 01:01:12.492269

# ceph tell osd.3 clear_shards_repaired

[待续]