Hyukjin Kwon created SPARK-61070:
------------------------------------
Summary: Stream the chunks of artifact uploads, and refuse class
files in addArtifacts, in the Rust client as PySpark does
Key: SPARK-61070
URL: https://issues.apache.org/jira/browse/SPARK-61070
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: 5.0.0
Reporter: Hyukjin Kwon
addArtifacts in the Rust Spark Connect client holds two copies of an upload
(the frames are built up front and cloned for each attempt), so uploading a 300
MB jar peaks at about 500 MB of memory where PySpark 4.3.0 peaks at about 170
MB, and the Python client uploads a .class file where PySpark raises
PySparkRuntimeError UNSUPPORTED_OPERATION.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
