[
https://issues.apache.org/jira/browse/HADOOP-9991?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14645437#comment-14645437
]
Steve Loughran commented on HADOOP-9991:
----------------------------------------
Hadoop only explicitly gets at log4j in some tests, so code-wise, excluding
those, switching to log4j v 2 is technically a matter of changing the mvn
version.
Except
* we don't know what happens downstream
* logging is essential to all services
* rolling logs of YARN services matters too.
For those reasons, it's hard to see enthusiasm for a change. Especially as
async logging has a price: when things fail the log may not be complete.
What is possible is to move the hadoop logging to slf4j off commons logging;
that's something that we'd like to see in new code, even though migrating old
code is TBD. Do that and switching to log4j or even logback can be done by
changing the slf4j dependency
> Fix up Hadoop Poms for enforced dependencies, roll up JARs to latest versions
> -----------------------------------------------------------------------------
>
> Key: HADOOP-9991
> URL: https://issues.apache.org/jira/browse/HADOOP-9991
> Project: Hadoop Common
> Issue Type: Improvement
> Components: build
> Affects Versions: 2.1.1-beta, 2.3.0
> Reporter: Steve Loughran
> Assignee: Steve Loughran
> Attachments: hadoop-9991-v1.txt
>
>
> If you try using Hadoop downstream with a classpath shared with HBase and
> Accumulo, you soon discover how messy the dependencies are.
> Hadoop's side of this problem is
> # not being up to date with some of the external releases of common JARs
> # not locking down/excluding inconsistent versions of artifacts provided down
> the dependency graph
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
