Yuli Fiterman created SPARK-27070:
-------------------------------------
Summary: DefaultPartitionCoalescer can lock up driver for hours
Key: SPARK-27070
URL: https://issues.apache.org/jira/browse/SPARK-27070
Project: Spark
Issue Type: Bug
Components: Spark Core
Affects Versions: 2.4.0, 2.3.2, 2.3.1
Reporter: Yuli Fiterman
We're running Spark on EMR reading large datasets from S3. When trying to
coalesce a UnionRDD of two large FileScanRDDs (each with a few million
partitions) into around 8k partitions the driver can stall for over an hour.
Profiler shows that over 90% of the time is spent in TimSort which is invoked
by `pickBin`. This seems like a very inefficient way to find the least occupied
PartitionGroup. IMO a better way would just using the `min` method on the
ArrayBuffer of `PartitionGroup`s
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
