Davis Zhang created HUDI-8491:
---------------------------------

             Summary: Fix concurrent schema evolution conflict bug
                 Key: HUDI-8491
                 URL: https://issues.apache.org/jira/browse/HUDI-8491
             Project: Apache Hudi
          Issue Type: Bug
            Reporter: Davis Zhang


For schema conflict resolution, previously we assume all instances with type 
COMMIT or DELTA_COMMIT comes with a valid schema. This is not true as there can 
be cases where hoodie streamer makes empty insert commits just for purpose of 
saving some commit metadata like the checkpoint offset while no data is written 
to the hudi table (maybe all messages fetched from a data source like kafka 
turns out to be not useful in that batch). In such a case, the commit metadata 
does not necessarily comes with a valid table schema ("schema" field is 
""null"")

The change accommodates for this corner case.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to