公司搬家，所有服务器的ip改变。对ceph服务器配置好ip后启动，发现monitor进程启动失败，monitor进程总是试图绑定到以前的ip地址，那当然不可能成功了。开始以为服务器的ip设置有问题，在改变hostname、ceph.conf等方法无果后，逐步分析发现，是monmap中的ip地址还是以前的ip，ceph通过读取monmap来启动monitor进程，所以需要修改monmap。方法如下：

#Add the new monitor locations
# monmaptool --create --add mon0 192.168.32.2:6789 --add osd1 192.168.32.3:6789   --add osd2 192.168.32.4:6789 --fsid 61a520db-317b-41f1-9752-30cedc5ffb9a   --clobber monmap

#Retrieve the monitor map
# ceph mon getmap -o monmap.bin

#Check new contents
# monmaptool --print monmap.bin

#Inject the monmap
# ceph-mon -i mon0 --inject-monmap monmap.bin
# ceph-mon -i osd1 --inject-monmap monmap.bin
# ceph-mon -i osd2 --inject-monmap monmap.bin

再启动monitor，一切正常。

但出现了上一篇文章中描述的一块osd盘挂掉的情况。查了一圈，只搜到ceph的官网上说是ceph的一个bug。无力修复，于是删掉这块osd，再重装：

# service ceph stop osd.4
#不必执行ceph osd crush remove osd.4
# ceph auth del osd.4
# ceph osd rm 4

# umount /cephmp1
# mkfs.xfs -f /dev/sdc
# mount /dev/sdc /cephmp1
#此处执行create无法正常安装osd
# ceph-deploy osd prepare osd2:/cephmp1:/dev/sdf1
# ceph-deploy osd activate osd2:/cephmp1:/dev/sdf1

[root@osd2 ~]# ceph -s
    cluster 61a520db-317b-41f1-9752-30cedc5ffb9a
     health HEALTH_WARN 9 pgs incomplete; 9 pgs stuck inactive; 9 pgs stuck unclean; 3 requests are blocked > 32 sec
     monmap e3: 3 mons at {mon0=192.168.32.2:6789/0,osd1=192.168.32.3:6789/0,osd2=192.168.32.4:6789/0}, election epoch 76, quorum 0,1,2 mon0,osd1,osd2
     osdmap e689: 6 osds: 6 up, 6 in
      pgmap v189608: 704 pgs, 5 pools, 34983 MB data, 8966 objects
            69349 MB used, 11104 GB / 11172 GB avail
                 695 active+clean
                   9 incomplete

出现了9个pg的incomplete状态。

[root@osd2 ~]# ceph health detail
HEALTH_WARN 9 pgs incomplete; 9 pgs stuck inactive; 9 pgs stuck unclean; 3 requests are blocked > 32 sec; 1 osds have slow requests
pg 5.95 is stuck inactive for 838842.634721, current state incomplete, last acting [1,4]
pg 5.66 is stuck inactive since forever, current state incomplete, last acting [4,0]
pg 5.de is stuck inactive for 808270.105968, current state incomplete, last acting [0,4]
pg 5.f5 is stuck inactive for 496137.708887, current state incomplete, last acting [0,4]
pg 5.11 is stuck inactive since forever, current state incomplete, last acting [4,1]
pg 5.30 is stuck inactive for 507062.828403, current state incomplete, last acting [0,4]
pg 5.bc is stuck inactive since forever, current state incomplete, last acting [4,1]
pg 5.a7 is stuck inactive for 499713.993372, current state incomplete, last acting [1,4]
pg 5.22 is stuck inactive for 496125.831204, current state incomplete, last acting [0,4]
pg 5.95 is stuck unclean for 838842.634796, current state incomplete, last acting [1,4]
pg 5.66 is stuck unclean since forever, current state incomplete, last acting [4,0]
pg 5.de is stuck unclean for 808270.106039, current state incomplete, last acting [0,4]
pg 5.f5 is stuck unclean for 496137.708958, current state incomplete, last acting [0,4]
pg 5.11 is stuck unclean since forever, current state incomplete, last acting [4,1]
pg 5.30 is stuck unclean for 507062.828475, current state incomplete, last acting [0,4]
pg 5.bc is stuck unclean since forever, current state incomplete, last acting [4,1]
pg 5.a7 is stuck unclean for 499713.993443, current state incomplete, last acting [1,4]
pg 5.22 is stuck unclean for 496125.831274, current state incomplete, last acting [0,4]
pg 5.de is incomplete, acting [0,4]
pg 5.bc is incomplete, acting [4,1]
pg 5.a7 is incomplete, acting [1,4]
pg 5.95 is incomplete, acting [1,4]
pg 5.66 is incomplete, acting [4,0]
pg 5.30 is incomplete, acting [0,4]
pg 5.22 is incomplete, acting [0,4]
pg 5.11 is incomplete, acting [4,1]
pg 5.f5 is incomplete, acting [0,4]
2 ops are blocked > 8388.61 sec
1 ops are blocked > 4194.3 sec
2 ops are blocked > 8388.61 sec on osd.0
1 ops are blocked > 4194.3 sec on osd.0
1 osds have slow requests

查了一圈无果。一个有同样遭遇的人的一段话：

I already tried "ceph pg repair 4.77", stop/start OSDs, "ceph osd lost", "ceph pg force_create_pg 4.77".
Most scary thing is "force_create_pg" does not work. At least it should be a way to wipe out a incomplete PG
without destroying a whole pool.

以上方法尝试了一下，都不行。暂时无法解决，感觉有点坑。

PS：常用pg操作

[root@osd2 ~]# ceph pg map 5.de
osdmap e689 pg 5.de (5.de) -> up [0,4] acting [0,4]
[root@osd2 ~]# ceph pg 5.de query
[root@osd2 ~]# ceph pg scrub 5.de
instructing pg 5.de on osd.0 to scrub
[root@osd2 ~]# ceph pg 5.de mark_unfound_lost revert
pg has no unfound objects
#ceph pg dump_stuck stale
#ceph pg dump_stuck inactive
#ceph pg dump_stuck unclean
[root@osd2 ~]# ceph osd lost 1
Error EPERM: are you SURE?  this might mean real, permanent data loss.  pass --yes-i-really-mean-it if you really do.
[root@osd2 ~]# 
[root@osd2 ~]# ceph osd lost 4 --yes-i-really-mean-it
osd.4 is not down or doesn‘t exist
[root@osd2 ~]# service ceph stop osd.4
=== osd.4 === 
Stopping Ceph osd.4 on osd2...kill 22287...kill 22287...done
[root@osd2 ~]# ceph osd lost 4 --yes-i-really-mean-it
marked osd lost in epoch 690
[root@osd1 mnt]# ceph pg repair 5.de
instructing pg 5.de on osd.0 to repair
[root@osd1 mnt]# ceph pg repair 5.de
instructing pg 5.de on osd.0 to repair

ip改变引起的ceph monitor异常及osd盘崩溃的总结