[
https://issues.apache.org/jira/browse/SPARK-41386?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17644141#comment-17644141
]
Zhe Dong commented on SPARK-41386:
----------------------------------
{noformat}
if (mapStats.isEmpty ||
mapStats.get.bytesByPartitionId.forall(_ <= advisorySize && _ >=
advisorySize * smallPartitionFactor )) {
return shuffle
}
--------------------------------------------------------------------------------
if (bytes > targetSize) {
...
} else if ( bytes < targetSize * smallPartitionFactor ){
CoalescedPartitionSpec(reduceIndex, reduceIndex + 1, bytes) :: Nil
}else {
return shuffle // dummy
}{noformat}
> There are some small files when using rebalance(column)
> -------------------------------------------------------
>
> Key: SPARK-41386
> URL: https://issues.apache.org/jira/browse/SPARK-41386
> Project: Spark
> Issue Type: Improvement
> Components: SQL
> Affects Versions: 3.4.0
> Reporter: Zhe Dong
> Priority: Minor
>
> *Problem ( REBALANCE(column)* {*}){*}:
> SparkSession config:
> {noformat}
> config("spark.sql.adaptive.optimizeSkewsInRebalancePartitions.enabled",
> "true")
> config("spark.sql.adaptive.advisoryPartitionSizeInBytes", "20m")
> config("spark.sql.adaptive.rebalancePartitionsSmallPartitionFactor",
> "0.5"){noformat}
> so, we except that files size should be bigger than 20m*0.5=10m at least.
> but in fact , we got some small files like the following:
> {noformat}
> -rw-r--r-- 1 jp28948 staff 12.1 M 2022-12-07 13:13
> .../part-00000-1ece1aae-f4f6-47ac-abe2-170ccb61f60e.c000.snappy.parquet
> -rw-r--r-- 1 jp28948 staff 12.1 M 2022-12-07 13:13
> .../part-00001-1ece1aae-f4f6-47ac-abe2-170ccb61f60e.c000.snappy.parquet
> -rw-r--r-- 1 jp28948 staff 12.1 M 2022-12-07 13:13
> .../part-00002-1ece1aae-f4f6-47ac-abe2-170ccb61f60e.c000.snappy.parquet
> -rw-r--r-- 1 jp28948 staff 12.1 M 2022-12-07 13:13
> .../part-00003-1ece1aae-f4f6-47ac-abe2-170ccb61f60e.c000.snappy.parquet
> -rw-r--r-- 1 jp28948 staff 9.1 M 2022-12-07 13:13
> .../part-00004-1ece1aae-f4f6-47ac-abe2-170ccb61f60e.c000.snappy.parquet
> -rw-r--r-- 1 jp28948 staff 3.0 M 2022-12-07 13:13
> .../part-00005-1ece1aae-f4f6-47ac-abe2-170ccb61f60e.c000.snappy.parquet{noformat}
> 9.1 M and 3.0 M is smaller than 10M. we have to handle these small files in
> another way.
--
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
