Big fall of performance in integrating Apache with Tomcat.
Following manual suggestions, I've integrated Tomcat 3.2 with Apache, using mod_jk, in order to have a more robust and fast product to manage the static html pages.
But, testing the new architecture with Jmeter, I've found out that the average time in accessing both dynamic pages (managed by my JSP application) and static pages had grown up to about 5 times (the comparison was made respectively with Tomcat and Apache, both as stand alone servers). I'm very surprised with this result as it makes futile the highly recommended integration of Apache and Tomcat.
I wonder if I'm making some mistake in configuration (as more probable) or if this result is confirmed in 'literature'. In this latter case, which solution do you suggest? Tomcat4.0 integration with Apache is more performant? Or I have to use Tomcat also for the static pages (I've seen that this doesn't affect that much the response times of static pages)?
The problem raises up from a bad configuration suggestion present in the official documentation of Tomcat 3.2 (maybe in other releases too).
In httpd.conf the log level for mod_jk (the exact name of the parameter to look for is JkLogLevel) does notexpect a value 'warn' as it is suggested. You'd better put it to 'error', for example. Otherwise this mistake causes the flushing of the syntax error on the disk at run-time, and as a consequence of this, all response times grow up.