How to understand the status code

[cigarl]

Hi,all

When I tried to fix bugs, I found that i am confused about the rules of the status code. For the same problem, the status code returned to the client may be different, and there is no clear hierarchy.

For example, When WAL failed, we will return WRITE_PROCESS_ERROR in insertRow, and INTERNAL_SERVER_ERROR in insertTablet,and i can't find the hierarchy between them.

Just like the directory structure of code, we need a clear hierarchy of response to client. Details of response in BACTH_INSERT should be in a same rank with the result of INSERT.

IMO, we need three levels at least, ATOMIC,BATCH and SERVER. The status of ATOMIC should belong to BATCH.And status codes of SERVER should be carefully returned to the client,that means there may be a serious problem in server,just like INTERNAL_SERVER_ERROR and READ_ONLY_SYSTEM_ERROR etc.

Of course, we might also need the cluster level.

Maybe, it needs everyone's opinions and ideas : )

[jixuan1989]

Hi,

IMO,
if a write (no matter what operation is, either insertRow or insertTablet) fails
because of users' invalid data, we need to return error codes like 401 (SQL parser error), 601 (no permission), 413 (write reject).

If the server has something wrong in the writing process, 412 (write process error) is returned.
If the server has something wrong that we do not know, 500 (server error) is returned.