Details
-
Bug
-
Status: Resolved
-
Minor
-
Resolution: Duplicate
-
1.5.1
-
None
-
None
Description
Noticed that slowly, over the course of a day or two, heap memory usage on a long running spark process increased monotonically.
After doing a heap dump and examining in jvisualvm, saw there were over 15M org.apache.spark.sql.execution.ui.SQLTaskMetrics objects allocated, taking over 500MB.
Attachments
Issue Links
- duplicates
-
SPARK-11126 A memory leak in SQLListener._stageIdToStageMetrics
- Resolved