MongoDb cannot be synchronized, probably because the server time is inconsistent

Today, I found image errors stored in MongoDb GridFS. I checked the logs and found many logs similar to the following:

Sat Mar 31 16:10:45 [conn42882] ns: fangdd. fs. chunks cocould not initialize cursor should SS all shards because: ns: mydb. fs. chunks ClusteredCursor: query @ shard2/192.168.222.201: 10021,192.168 .222.202: 10024,192.168 .222.202: 10022,192.168 .222.201: 10023 attempt: 3

I had to use ps ax | grep mongo to find out each mongo and mongos process, kill-2 nnn one by one, and then start. In addition, the servers involved in the Cluster also perform the same operation.

After the restart is completed, the following errors are found:

Sat Mar 31 17:23:10 [Balancer] caught exception while doing balance: error checking clock skew of cluster 192.168.222.201: 20000,192.168 .222.201: 2001ew of the cluster 192.168.222.201: 20000,192.168 .222.201: 20010,192.168 .222.202: 20000 is too far out of bounds to allow distributed locking.

Later, it was found that the time of the server was not exactly the same, and one server was slow for more than 70 seconds.

Sudo ntpdate stdtime.sinica.edu.tw is normal after the school is completed.