Enrico Minack created SPARK-50417:
-------------------------------------
Summary: Limit number of subdirectories that FallbackStorage
creates per shuffle
Key: SPARK-50417
URL: https://issues.apache.org/jira/browse/SPARK-50417
Project: Spark
Issue Type: Improvement
Components: SQL
Affects Versions: 4.0.0
Reporter: Enrico Minack
The {{FallbackStorage}} copies shuffle data during executor decommissioning to
a distributed or cloud storage like S3 or HDFS. In
https://github.com/apache/spark/pull/34762, a hash has been added to the path
of a file in order to reduce the number of files per directory (prefix in S3
terms). This creates as many directories per shuffle as files are transferred,
while each directory contains a single file.
While this might be useful for S3, it may pose challenges for other
filesystems. A shuffle of 100,000 partitions creates 100,000 directories, each
containing a single file. The number of directories should be configurable to
be able to adjust this behavior for the specific filesystem used.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
