[
https://issues.apache.org/jira/browse/HADOOP-18207?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17702028#comment-17702028
]
Viraj Jasani commented on HADOOP-18207:
---------------------------------------
Basically the idea is to put all the code that ever needs to directly touch
log4j APIs under a new module so that future upgrades also become easier to
deal with, as opposed to what we are currently going through.
Certain things are not directly achieved by SLF4J like setting log level
dynamically (we have dedicated servlet to make this happen). We also still have
one custom appender that we will have to manage with log4j2 as well AFAIK, it's
used by yarn: Log4jWarningErrorMetricsAppender.
We have several tests that still use log4j APIs directly in order to add
appenders dynamically, something that can also be moved to new module.
If we could do this first before the actual log4j2 migration, this would be
quite structured and well maintained. We would have only one module that deals
with log4j APIs and others don't have to.
> Introduce a hadoop-logging module
> ---------------------------------
>
> Key: HADOOP-18207
> URL: https://issues.apache.org/jira/browse/HADOOP-18207
> Project: Hadoop Common
> Issue Type: Sub-task
> Reporter: Duo Zhang
> Assignee: Viraj Jasani
> Priority: Major
>
> There are several goals here:
> 1. Provide the ability to change log level, get log level, etc.
> 2. Place all the appender implementation(?)
> 3. Hide the real logging implementation.
> 4. Later we could remove all the log4j references in other hadoop module.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
