For example, my mongodb shard cluster originally had 2 shards. After opening the shards for the test.user set, insert the data into the test.user. Observation shows that the data has been evenly distributed on the original 2 shards. Then, a new shard has been added at the later stage. Then insert the data into the test.user again. It is found that only the original 2 shards still have data, but there is no data on the new shard!

If other sets such as test.role are opened after adding a slice, the inserted data can be evenly distributed over the 3 slices!