vanzin commented on a change in pull request #25863:
[SPARK-28945][SPARK-29037][CORE][SQL] Fix the issue that spark gives duplicate
result and support concurrent file source write operations write to different
partitions in the same table.
URL: https://github.com/apache/spark/pull/25863#discussion_r328874886
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/InsertIntoHadoopFsRelationCommand.scala
##########
@@ -263,4 +280,107 @@ case class InsertIntoHadoopFsRelationCommand(
}
}.toMap
}
+
+ /**
+ * Detect the conflict when there are several InsertIntoHadoopFsRelation
operations
+ * write to same partition in the same table or a non-partitioned table
concurrently.
+ */
+ private def detectConflict(
Review comment:
All this is happening before the `committer.setupJob` call, right?
Doesn't that mean that two applications may concurrently pass this test, and
then mess up each other when they actually start running the job?
It feels like there need to be a fancier protocol here to define who can
continue past this point, which would actually involve writing to the
underlying file system. e.g. by creating the needed staging directory(ies)
before doing this check (throwing an error if the directory already exists).
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
