Vitali Makarevich created HUDI-7914:
---------------------------------------
Summary: Incorrect schema produced in DELETE_PARTITION
replacecommit
Key: HUDI-7914
URL: https://issues.apache.org/jira/browse/HUDI-7914
Project: Apache Hudi
Issue Type: Bug
Reporter: Vitali Makarevich
in the current scenario delete_partitions produces {{replacecommit}} with
internal fields - like {{{}_hoodie_file_name{}}}, while e.g. normal {{commit}}
produces schema without such fields.
This leads to unexpected behavior when the {{replacecommit}} is the last on the
commitline,
e.g. [#10258|https://github.com/apache/hudi/issues/10258]
[#10533|https://github.com/apache/hudi/issues/10533]
and e.g. metadata sync things, or any other potential write will take incorrect
schema - and in the best case will fail because fields are duplicated, in the
worst cases can lead to dataloss.
The problem introduced here [https://github.com/apache/hudi/pull/5610/files]
And for other operations like {{delete}} the same approach used as I use now.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
