Github user sryza commented on the pull request:
https://github.com/apache/spark/pull/148#issuecomment-37752740
Currently, Spark doesn't ship a log4j.properties. It uses the
log4j.properties that comes from Hadoop. This log4j.properties is meant for
Hadoop services, not YARN containers. container-log4j.properties is a log4j
configuration that Hadoop provides for containers to use.
Without this change, is there a way to control executor logging without
turning on the config option that puts the user classpath first and then
including a log4j.properties in the app jar? This doesn't seem like a good way
to configure logging to me. Though with the change, users can still control
the change by doing the same, but instead including a file named
container-log4j.properties in their app jar.
I would be in favor of providing better ways for users to configure logging
for executors, but, if I understand correctly, this change isn't making things
worse.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
