操作步骤
1. Monitor操作，将OSD的状态设置Out
2. 登录对应OSD节点 ，停止OSD守护进程
3. Crushmap中删除OSD
4. 删除OSD的认证信息
5. 登录对应OSD节点 ，取消挂载文件系统
注意事项
停止OSD和删除Crushmao的信息，都将导致PG的重新收敛，所以最好每一个步骤都确保当前ceph状态是OK才进行下一步。
######参考脚本
#!/bin/sh
####judge state of ceph
judge() {
while true
do
sleep 5
ceph -s |grep HEALTH_OK
if [[ $? -eq 0 ]]
then
break
else
sleep 60
fi
done
}
####rdel osd from host14
while true
do
for host in `seq 31 39`
do
echo "now.... ,del osd.$host from host14" >>stat.txt
ceph osd out $host
judge
ssh 192.168.220.14 "/etc/init.d/ceph stop osd.$host"
judge
ceph osd crush remove osd.$host
judge
ceph auth del osd.$host
judge
ceph osd rm osd.$host
judge
ssh 192.168.220.14 "umount /var/lib/ceph/osd/ceph-$host"
judge
echo "del osd.$host from host14 suceess!!!!" >>stat.txt
done
break
done
######del osd from host16
while true
do
for host in `seq 40 45`
do
echo "now.... ,del osd.$host from host16" >>stat.txt
ceph osd out $host
judge
ssh 192.168.220.16 "/etc/init.d/ceph stop osd.$host"
judge
ceph osd crush remove osd.$host
judge
ceph auth del osd.$host
judge
ceph osd rm osd.$host
judge
ssh 192.168.220.16 "umount /var/lib/ceph/osd/ceph-$host"
judge
echo "del osd.$host from host16 suceess!!!!" >>stat.txt
done
break
done
Reference
-------他日江湖相逢 再当杯酒言欢-------
