[jira] [Resolved] (GEODE-8875) Geode is spending a lot of time in log4j classpath scanning on startup

["Dan Smith (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/GEODE-8875?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Dan Smith resolved GEODE-8875.
------------------------------
    Fix Version/s: 1.14.0
         Assignee: Dan Smith
       Resolution: Fixed

> Geode is spending a lot of time in log4j classpath scanning on startup
> ----------------------------------------------------------------------
>
>                 Key: GEODE-8875
>                 URL: https://issues.apache.org/jira/browse/GEODE-8875
>             Project: Geode
>          Issue Type: Improvement
>          Components: logging
>            Reporter: Dan Smith
>            Assignee: Dan Smith
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 1.14.0
>
>
> When running gfsh, it takes a long time even to run the `gfsh help` command. For example, with our docker image it takes 10 seconds just to display the help from
> {code}
> docker run apachegeode/geode:latest time gfsh help
> {code}
> Doing some CPU sampling, it looks like the majority of the time is in log4j2 scanning for plugins.
> {noformat}
> TRACE 300879: (thread=200001)
>         java.lang.ClassLoader.defineClass1(ClassLoader.java:Unknown line)
>         java.lang.ClassLoader.defineClass(ClassLoader.java:763)
>         java.security.SecureClassLoader.defineClass(SecureClassLoader.java:142)
>         java.net.URLClassLoader.defineClass(URLClassLoader.java:468)
>         java.net.URLClassLoader.access$100(URLClassLoader.java:74)
>         java.net.URLClassLoader$1.run(URLClassLoader.java:369)
>         java.net.URLClassLoader$1.run(URLClassLoader.java:363)
>         java.security.AccessController.doPrivileged(AccessController.java:Unknown line)
>         java.net.URLClassLoader.findClass(URLClassLoader.java:362)
>         java.lang.ClassLoader.loadClass(ClassLoader.java:424)
>         sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:349)
>         java.lang.ClassLoader.loadClass(ClassLoader.java:357)
>         org.apache.logging.log4j.core.config.plugins.util.ResolverUtil.addIfMatching(ResolverUtil.java:449)
>         org.apache.logging.log4j.core.config.plugins.util.ResolverUtil.loadImplementationsInJar(ResolverUtil.java:351)
>         org.apache.logging.log4j.core.config.plugins.util.ResolverUtil.findInPackage(ResolverUtil.java:233)
>         org.apache.logging.log4j.core.config.plugins.util.PluginRegistry.loadFromPackage(PluginRegistry.java:221)
>         org.apache.logging.log4j.core.config.plugins.util.PluginManager.collectPlugins(PluginManager.java:152)
>         org.apache.logging.log4j.core.config.AbstractConfiguration.initialize(AbstractConfiguration.java:224)
>         org.apache.logging.log4j.core.config.AbstractConfiguration.start(AbstractConfiguration.java:288)
>       org.apache.logging.log4j.core.LoggerContext.setConfiguration(LoggerContext.java:579)
> {noformat}
> After further investigation, this scanning appears to be caused by this line in our log4j2-cli.xml file, which sets the packages to be "org.apache.geode" This causes log4j2 to scan all of our jar files.
> {code}
> <Configuration status="WARN" shutdownHook="disable" packages="org.apache.geode">
> {code}
> This scanning is completely unnecessary because log4j already has an annotation processor which generates a file that log4j2 can find without any scanning.
> This probably affects geode server startup, not just gfsh, because log4j2.xml has this setting as well.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)