Hi all, I am running spark standalone (local[*]), and have tried to cut back on some of the logging noise from the framework by editing log4j.properties in spark/conf. The following lines are working as expected:
log4j.logger.org.apache.spark=WARN log4j.logger.org.apache.spark.storage.BlockManager=ERROR (I've even guaranteed that it's definitely using my configuration by adding a prefix to the conversion pattern). However, I am still getting log messages at INFO from classes like: org.apache.spark.Logging$class (should be covered by the org.apache.spark setting) kafka.utils.Logging$class (when I add a similar setting for kafka.utils) I suspect it's because these are inner classes. It still happens even when I go up a level and add configurations like "log4j.logger.org=WARN". Is this a known bug in log4j? Is there any known way to suppress these, ideally through configuration rather than programmatically? Many thanks