[SPARK-24787] Events being dropped at an alarming rate due to hsync being slow for eventLogging

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 2.3.0, 2.3.1
• Fix Version/s: 2.4.0
• Component/s: Spark Core, Web UI
• Assignee: Devaraj Kavali
• Reporter: Sanket Reddy
• Created: 11/Jul/18 20:47
• Updated: 29/Oct/18 00:58
• Resolved: 25/Oct/18 21:05

Description

https://github.com/apache/spark/pull/16924/files updates the length of the inprogress files allowing history server being responsive.

Although we have a production job that has 60000 tasks per stage and due to hsync being slow it starts dropping events and the history server has wrong stats due to events being dropped.

A viable solution is not to make it sync very frequently or make it configurable.

Issue Links

• duplicates: SPARK-25645 Add provision to disable EventLoggingListener default flush/hsync/hflush for all events
• is caused by: SPARK-19531 History server doesn't refresh jobs for long-life apps like thriftserver
• links to: [Github] Pull Request #22752 (devaraj-kavali)
• links to: [Github] Pull Request #22752 (devaraj-kavali)