ceph修复pg inconsistent（ scrub errors）

异常情况

1、收到异常情况如下:

OSD_SCRUB_ERRORS 12 scrub errors
PG_DAMAGED Possible data damage: 1 pg inconsistentpg 6.d is active+remapped+inconsistent+backfill_wait, acting [5,7,4]

2、查看详细信息

登录后复制

#ceph health detail
HEALTH_ERR 12 scrub errors; Possible data damage: 1 pg inconsistent
OSD_SCRUB_ERRORS 12 scrub errors
PG_DAMAGED Possible data damage: 1 pg inconsistentpg 6.d is active+remapped+inconsistent+backfill_wait, acting [5,7,4]

2、预处理办法

集群出现pg状态变为 inconsistent 状态，inconsistent状态是pg通过scrub检测到某个或者某些对象在副本之间出现了不一致

解决方法：

1.找到pgid

ceph pg dump | grep inconsistent

2.通过repair修复对应的pg

ceph pg repair pgid

完成
但是经过观察，并不能解决。

3、我的处理过程

找出异常的 pg，然后到对应的osd所在主机上修复。

[root@intelligence-b-10 ~]#  ceph osd find 5
{"osd": 5,"ip": "10.21.230.91:6802/2032","crush_location": {"host": "intelligence-b-12","root": "default"}
}
[root@intelligence-b-10 ~]#  ceph osd find 7
{"osd": 7,"ip": "10.21.230.92:6801/2059","crush_location": {"host": "intelligence-b-13","root": "default"}
}
[root@intelligence-b-10 ~]#  ceph osd find 4
{"osd": 4,"ip": "10.21.230.90:6800/2084","crush_location": {"host": "intelligence-b-11","root": "default"}
}

这里表示是主机intelligence-b-12
然后到 进行修复

1、停止osd

systemctl stop [e-mail]

2、刷入日志

ceph-osd -i 5 --flush-journal

3、启动osd

systemctl start [e-mail]

4、修复（一般不需要）

ceph pg repair 6.d

5、查看ceph的状态

ceph -s