Gera Shegalov created MAPREDUCE-6280:
----------------------------------------
Summary: Reject directory vs file path conflict resolution in
FileOutputCommitter
Key: MAPREDUCE-6280
URL: https://issues.apache.org/jira/browse/MAPREDUCE-6280
Project: Hadoop Map/Reduce
Issue Type: Bug
Components: mrv2
Affects Versions: 2.6.0
Reporter: Gera Shegalov
If one task commits a directory {{foo}}, and then another task commits file
{{foo}}, the directory {{foo}} with potentially many files will be wiped out.
While this is a very unlikely scenario, due to tasks being homogeneous in
nature, it's so much more important to alert the user by failing the commit.
This came up in [~jlowe]'s review for MAPREDUCE-6275 and seems to be [the
behavior in
branch-1|https://github.com/apache/hadoop/blob/branch-1/src/mapred/org/apache/hadoop/mapreduce/lib/output/FileOutputCommitter.java#L198]
as well.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)