[
https://issues.apache.org/jira/browse/SPARK-2654?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14964586#comment-14964586
]
Jeff Zhang commented on SPARK-2654:
-----------------------------------
[~davies] I think currently spark-core also don't have logging level control
through command line argument except through log4j.properties. Although
implementation for controlling logging level in spark-core and pyspark will be
different, but the configuration should be same.
Here's my initial thinking about it:
* use "spark.driver.log.level" to control log level of driver and
"spark.executor.log.level" to control log level of "executor".
* "spark.driver.log.level" and "spark.executor.log.level" can be set in 2 ways
** Simple Configuration: just log level like INFO/DEBUG/ERROR/...
** Advanced Configuration: Simple Configuration followed by package level
configuration, like
"DEBUG;org.apache.spark.sql=INFO;org.apache.spark.shuffle=INFO"
I am not sure whether there's existing jira of logging level control for
spark-core, if you know that, please help link them together. I'd like to help
contribute this. Thanks
> Leveled logging in PySpark
> --------------------------
>
> Key: SPARK-2654
> URL: https://issues.apache.org/jira/browse/SPARK-2654
> Project: Spark
> Issue Type: Improvement
> Components: PySpark
> Reporter: Davies Liu
>
> Add more leveled logging in PySpark, the logging level should be easy
> controlled by configuration and command line arguments.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
