Consistent File System Model of hadoop

It is equivalent to the visibility of Java synchronization. After a block is fully written, the data stored in it is visible. Even if the file description is visible, its length may be 0. even if the data has been actually written to the block.

In most cases, this does not affect our file requirements. For files stored on hadoop, we do not use the content in the file as a signal mark.

It means you do not understand his design purpose.

However, if you have such abnormal requirements, you can still call sync to force data on unfilled blocks to be visible in the file system.

Generally, you can call the close method to implicitly call sync after the file write operation is complete to publish the file visibility.