cfmcgrady commented on code in PR #2482:
URL: https://github.com/apache/celeborn/pull/2482#discussion_r1591166149
##########
client-spark/spark-3/src/main/java/org/apache/spark/shuffle/celeborn/SparkShuffleManager.java:
##########
@@ -420,6 +424,20 @@ private int executorCores(SparkConf conf) {
}
}
+ private void checkUserClassPathFirst() {
+ if ((Boolean) conf.get(package$.MODULE$.EXECUTOR_USER_CLASS_PATH_FIRST()))
{
Review Comment:
Whatmore, can you explain why the configuration
`spark.executor.userClassPathFirst=false` is required for Spark version >=
4.0.0? This means that even if we place the Celeborn client JAR in the
`$SPARK_HOME/jars` folder and set `spark.executor.userClassPathFirst=true`, the
`SparkShuffleManager` still falls back to the vanilla Spark
`SortShuffleManager`?
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
