[SPARK-38881] PySpark Kinesis Streaming should expose metricsLevel CloudWatch config that is already supported in the Scala/Java APIs

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 3.2.1
• Fix Version/s: 3.4.0
• Component/s: DStreams, Input/Output, PySpark
• Assignee: Mark Khaitman
• Reporter: Mark Khaitman
• Created: 12/Apr/22 18:44
• Updated: 17/Apr/22 02:30
• Resolved: 17/Apr/22 02:30

Description

This relates to https://issues.apache.org/jira/browse/SPARK-27420 which was merged as part of Spark 3.0.0

This change is desirable as it further exposes the metricsLevel config parameter that was added for the Scala/Java Spark APIs when working with the Kinesis Streaming integration, and makes it available to the PySpark API as well.

This change passes all tests, and local testing was done with a development Kinesis stream in AWS, in order to confirm that metrics were no longer being reported to CloudWatch after specifying MetricsLevel.NONE in the PySpark Kinesis streaming context creation, and also worked as it does today when leaving the MetricsLevel parameter out, which would result in a default of DETAILED, with CloudWatch metrics appearing again.

https://github.com/apache/spark/pull/36201

 

Issue Links

• links to: [Github] Pull Request #36201 (mkman84)