Tag Profiler Plugin

This plugin is aimed at plugin developers. It allows you to profile different sections of jelly tags in order to identify response bottlenecks.

By default this plugin will profile all requests. To profile a specific part of a page just surround that part of the page with a <p:profile name="...">...</p:profile> tag, e.g.

<j:jelly xmlns:j="jelly:core" xmlns:st="jelly:stapler" xmlns:l="/lib/layout"
         xmlns:f="/lib/form" xmlns\:p="/lib/profile">
  <l:layout title="${it.displayName}">
    <p:profile name="side-panel">
      <st:include page="sidepanel.jelly"/
    </p:profile>
    <p:profile name="main-panel">
      ...
    </p:profile>
  </l:layout>
</j:jelly>

You can also use the profiler from within your code, e.g.

import org.jenkinsci.plugins.tagprofiler.ProfileStats;
...
    ProfileStats.enter("critical");
    try {
        ...
    } finally {
        ProfileStats.leave();
    }

The profiling results are available from the /tag-profiler root action. They look something like this:
The Reset button will reset the current measurements. The ± value is the standard error of the mean.

Tips & Tricks

• When using hpi:run, be aware that the jelly pages are checked for modification on every access, so performance can appear worse than when running in stock Jenkins
• This kind of profiling is only really intended for those cases where a regular profiler will not help, namely when profiling within jelly tags. Very often there are some simple changes in jelly that will result in a major performance boost.
• If you are repeatedly evaluating the same EL fragment, consider caching that fragment with <j:set var="" value=""/> this can have a major performance boost where the EL fragment is time-costly to evaluate.

Version History

Version 0.2 (2013-02-15)

• Added configuration options to disable profiling and to turn off the request URI categorization of profile segments

Version 0.1 (2013-02-15)

• Initial release