[[379685]]

 今天檢查ceph集群，發現有pg丟失，于是就有了本文~~~

1.查看集群狀態

[root@k8snode001 ~]# ceph health detail 
HEALTH_ERR 1/973013 objects unfound (0.000%); 17 scrub errors; Possible data damage: 1 pg recovery_unfound, 8 pgs inconsistent, 1 pg repair; Degraded data redundancy: 1/2919039 objects degraded (0.000%), 1 pg degraded 
OBJECT_UNFOUND 1/973013 objects unfound (0.000%) 
    pg 2.2b has 1 unfound objects 
OSD_SCRUB_ERRORS 17 scrub errors 
PG_DAMAGED Possible data damage: 1 pg recovery_unfound, 8 pgs inconsistent, 1 pg repair 
    pg 2.2b is active+recovery_unfound+degraded, acting [14,22,4], 1 unfound 
    pg 2.44 is active+clean+inconsistent, acting [14,8,21] 
    pg 2.73 is active+clean+inconsistent, acting [25,14,8] 
    pg 2.80 is active+clean+scrubbing+deep+inconsistent+repair, acting [4,8,14] 
    pg 2.83 is active+clean+inconsistent, acting [14,13,6] 
    pg 2.ae is active+clean+inconsistent, acting [14,3,2] 
    pg 2.c4 is active+clean+inconsistent, acting [8,21,14] 
    pg 2.da is active+clean+inconsistent, acting [23,14,15] 
    pg 2.fa is active+clean+inconsistent, acting [14,23,25] 
PG_DEGRADED Degraded data redundancy: 1/2919039 objects degraded (0.000%), 1 pg degraded 
    pg 2.2b is active+recovery_unfound+degraded, acting [14,22,4], 1 unfound 

 從輸出發現pg 2.2b is active+recovery_unfound+degraded, acting [14,22,4], 1 unfound

現在我們來查看pg 2.2b，看看這個pg的想想信息。

[root@k8snode001 ~]# ceph pg dump_json pools    |grep 2.2b 
dumped all 
2.2b       2487                  1        1         0       1  9533198403 3048     3048                active+recovery_unfound+degraded 2020-07-23 08:56:07.669903  10373'5448370  10373:7312614  [14,22,4]         14  [14,22,4]             14  10371'5437258 2020-07-23 08:56:06.637012   10371'5437258 2020-07-23 08:56:06.637012             0 

 可以看到它現在只有一個副本

2.查看pg map

[root@k8snode001 ~]# ceph pg map 2.2b 
osdmap e10373 pg 2.2b (2.2b) -> up [14,22,4] acting [14,22,4] 

 從pg map可以看出，pg 2.2b分布到osd [14,22,4]上

3.查看存儲池狀態

[root@k8snode001 ~]# ceph osd pool stats k8s-1 
pool k8s-1 id 2 
  1/1955664 objects degraded (0.000%) 
  1/651888 objects unfound (0.000%) 
  client io 271 KiB/s wr, 0 op/s rd, 52 op/s wr 
 
[root@k8snode001 ~]# ceph osd pool ls detail|grep k8s-1 
pool 2 'k8s-1' replicated size 3 min_size 1 crush_rule 0 object_hash rjenkins pg_num 256 pgp_num 256 last_change 88 flags hashpspool,selfmanaged_snaps stripe_width 0 application rbd 

 4.嘗試恢復pg 2.2b丟失地塊

[root@k8snode001 ~]# ceph pg repair 2.2b 

如果一直修復不成功，可以查看卡住PG的具體信息，主要關注recovery_state，命令如下

[root@k8snode001 ~]# ceph pg 2.2b  query 
{ 
    "...... 
    "recovery_state": [ 
        { 
            "name": "Started/Primary/Active", 
            "enter_time": "2020-07-21 14:17:05.855923", 
            "might_have_unfound": [], 
            "recovery_progress": { 
                "backfill_targets": [], 
                "waiting_on_backfill": [], 
                "last_backfill_started": "MIN", 
                "backfill_info": { 
                    "begin": "MIN", 
                    "end": "MIN", 
                    "objects": [] 
                }, 
                "peer_backfill_info": [], 
                "backfills_in_flight": [], 
                "recovering": [], 
                "pg_backend": { 
                    "pull_from_peer": [], 
                    "pushing": [] 
                } 
            }, 
            "scrub": { 
                "scrubber.epoch_start": "10370", 
                "scrubber.active": false, 
                "scrubber.state": "INACTIVE", 
                "scrubber.start": "MIN", 
                "scrubber.end": "MIN", 
                "scrubber.max_end": "MIN", 
                "scrubber.subset_last_update": "0'0", 
                "scrubber.deep": false, 
                "scrubber.waiting_on_whom": [] 
            } 
        }, 
        { 
            "name": "Started", 
            "enter_time": "2020-07-21 14:17:04.814061" 
        } 
    ], 
    "agent_state": {} 
} 

 如果repair修復不了;兩種解決方案，回退舊版或者直接刪除

5.解決方案

回退舊版 
[root@k8snode001 ~]# ceph pg  2.2b  mark_unfound_lost revert 
直接刪除 
[root@k8snode001 ~]# ceph pg  2.2b  mark_unfound_lost delete 

 6.驗證

我這里直接刪除了，然后ceph集群重建pg,稍等會再看，pg狀態變為active+clean

[root@k8snode001 ~]#  ceph pg  2.2b query 
{ 
    "state": "active+clean", 
    "snap_trimq": "[]", 
    "snap_trimq_len": 0, 
    "epoch": 11069, 
    "up": [ 
        12, 
        22, 
        4 
    ], 

 再次查看集群狀態

[root@k8snode001 ~]# ceph health detail 
HEALTH_OK 

 【編輯推薦】