I now have a true live example of how text matching can provide information on issues where a successful page is returned.
In this example, the TEST AGENT returned a Text Match Failed error, while 3 of the agents running the current production code said the page was a success.
How do I know that the TEST AGENT is right? Take a look at the byte count. For the successful pages, the byte count is in the 3,600-3,900 byte range; the page that had the Text Match failure only returned 1076 bytes. And three other measurements around that time reported the same approximate size, but reported successful page downloads.
If this Agent code shows continued success and robust behaviour, then I will push it into production on August 14.
Technorati Tags: GrabPERF, Agent+upgrade, Web+performance, text+match, content+match