[
https://issues.apache.org/jira/browse/LOG4J2-1318?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15213472#comment-15213472
]
Remko Popma commented on LOG4J2-1318:
-------------------------------------
Looking further, SLF4JLoggerContext is implemented a bit differently from the
other LoggerContext implementations (SimpleLoggerContext, log4j-core
LoggerContext and Log4jTaglibLoggerContext): the others all check if the
previously registered logger matches the MessageFactory by calling
{{AbstractLogger.checkMessageFactory(logger, messageFactory)}}, but
SLF4JLoggerContext does not.
Same question: is this a bug or is SLF4JLoggerContext special for some reason
and is this difference by design?
> LoggerContext#getLogger causes heavy GC overhead
> ------------------------------------------------
>
> Key: LOG4J2-1318
> URL: https://issues.apache.org/jira/browse/LOG4J2-1318
> Project: Log4j 2
> Issue Type: Bug
> Components: Core
> Affects Versions: 2.5
> Reporter: Rodrigo Merino
> Assignee: Remko Popma
> Labels: performance
>
> With the changes from LOG4J2-1180, when running performance testing
> scenarios, we are experiencing some throughput degradations due to an
> increase in the GC stall time.
> With the changes from LOG4J2-1180, the GC stall went from ~6% to ~11%, having
> an impact on the application of a reduction of ~32% of the operations
> throughput. The memory allocation rate both before and after the change is
> 4GB/s and 12GB/s respectively.
> In both cases, the relevant jvm configuration params were:
> {code}
> -XX:+PrintGCApplicationStoppedTime -XX:+PrintGCDetails -XX:+PrintGCDateStamps
> -XX:+PrintTenuringDistribution -XX:ErrorFile=%MHOME%/logs/err.log
> -Xloggc:%HOME%/logs/gc.log -XX:+HeapDumpOnOutOfMemoryError
> -XX:+AlwaysPreTouch -Xms2048m -XX:PermSize=256m -XX:MaxPermSize=256m
> -XX:+UseParNewGC -XX:NewSize=1024m -XX:+UseNUMA
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
