MongoDB机器未能正常关机，机器掉电重启等各种情况，都可能引起mongodb数据文件异常，然后下次启动的时候，会报错大概如下
517322 2018-01-03T22:20:45.939+0800 I NETWORK [Balancer] scoped connection to 10.249.3.12:20000,10.249.3.11:20000,10.249.3.11:20003 not being returned to the pool
517323 2018-01-03T22:20:45.939+0800 I SHARDING [Balancer] caught exception while doing balance: ReplicaSetMonitor no master found for set: shard1
517324 2018-01-03T22:20:46.540+0800 I NETWORK [LockPinger] SyncClusterConnection connecting to [10.249.3.12:20000]
517325 2018-01-03T22:20:46.541+0800 I NETWORK [LockPinger] SyncClusterConnection connecting to [10.249.3.11:20000]
517326 2018-01-03T22:20:46.541+0800 I NETWORK [LockPinger] SyncClusterConnection connecting to [10.249.3.11:20003]
517327 2018-01-03T22:20:49.748+0800 W SHARDING config servers 10.249.3.12:20000 and 10.249.3.11:20000 differ
517328 2018-01-03T22:20:49.749+0800 W SHARDING config servers 10.249.3.12:20000 and 10.249.3.11:20000 differ
517329 2018-01-03T22:20:49.751+0800 W SHARDING config servers 10.249.3.12:20000 and 10.249.3.11:20000 differ
517330 2018-01-03T22:20:49.752+0800 W SHARDING config servers 10.249.3.12:20000 and 10.249.3.11:20000 differ
重点在differ这里，两个分片不一致，导致不能启动，解决方法如下：
分别安全关闭10.249.3.11和12的MongoDB
/usr/local/mongodb/bin/mongo -host 127.0.0.1 -port 20000
> use admin;
> db.shutdownServer();
进入到3.11机器的对应数据目录，将文件整个打包，
进入到3.12机器的数据目录/data/mongodb/config20001，删除所有文件，记得备份，
最后scp 11.zip 10.249.3.12:/data/mongodb/config20001 ，然后解压，最后启动mongodb即可。