Ever notice that your new EAP implementation appears to be leaking threads? Thread dump pointing to AJP? Does your thread usage resemble the graphs on the top, continuously climbing?

threads

Fortunately, the solution is simple- JBoss and mod_cluster are too brain damaged to account for Apache HTTPD timing out threads and closing them down, so it leaves them having open forever rather than have a sane (or even insane) default timeout. To work around this, add the following to your system-properties in your host.xml, standalone.xml or domain.xml:

<system-properties>
 ...
 <property name="org.apache.coyote.ajp.DEFAULT_CONNECTION_TIMEOUT" value="600000"/>
 ...
 <system-properties>

Poof, after 10 minutes of idle time, JBoss kills the thread, resulting in the graphs on the bottom after a restart. Leave a comment if this was helpful.