[FX.php List] Apache anolmaly with Fx.php and FileMaker Web Publishing Engine

[William Vaughn]

Tenon Technical Support:

Your previous analysis of the atitle.PHP script went a long way toward 
helping me understand the problem we are having with server configuration.

We are running iTools 8.2,  Tenon PHP 5.1.1, Tenon Tomcat 5.5.9, FX.PHP, 
and FileMaker Web Publishing 8v3 on a Mac Mini G4 1.25GHz with 1GB RAM 
(Web Server).

A second server is running FileMaker Server Advanced 8v3 on a Mac G4 
2.0GHz with 1.5GB RAM (Database Server).

The Database Server is running FileMaker Server Advanced and hosts 
FileMaker databases containing information that will be used to power 
websites hosted on the Web Server. The Web Server hosts websites using 
Apache and mod_PHP to publish dynamic content from FileMaker. PHP 
powered websites use mod_PHP running FX.PHP to communicate with 
FileMaker. The Web Publishing Engine communicates with FileMaker Server 
running the Database Server, and makes xml output available to FX.PHP, 
which in turn processes the data into an HTML document served by Apache. 
Tenon Tomcat is disabled, as the only reason it is installed is to 
permit Apache to use the bundled mod_jk connector to communicate with 
the Web Publishing Engine.

I have a test database set up on the Database server with a few tables 
and a couple of relationships. In one table I have ~5000 records with 
four or five text fields: name, title, date, and description.  I 
designed a PHP page which displayed 500 of those records in a list. At 
this point I encountered first encountered the problem. When loading the 
page (it took two or three seconds) the Web Publishing Engine placed a 
significant load on the processor for two or three seconds while the 
page was loading. That was expected. What followed was not. One of  the 
Apache httpd processes then pegged the processor at 85%-95%, AFTER the 
page had loaded in the browser. This was quite confusing. So, I went to 
the Web Server and opened Safari. I loaded the page. Sure enough, the 
WPE (Web Publishing Engine) worked for about three seconds and produced 
the page. Apache did not so much as blink. I set my laptop down next to 
the KVM and loaded the same page. Lo and behold, the WPE worked for two 
or three seconds, and the page loaded. Then, once again, Apache pegged 
the processor for several seconds, AFTER the page had loaded. So I did 
some experimentation, and found that if I limited the page to about 130 
records the problem went away: 130 records, and the WPE engine worked 
for two seconds; 135 records and the WPE worked for the same two 
seconds, but this time Apache pegged the processor for five to ten 
seconds after the page had loaded. I experimented more and found that 
the number of records was not the only factor. Fewer records with more 
data, or, vice versa, more records with fewer data, would exhibit the 
aberrant behavior. There was always a distinct limit, which if reached, 
would cause Apache to go misbehave as described above.

I have also run the script with the PHP CLI and , regardless of the 
number of record the page is set for,it always behaves more or less on a 
linear scale. (double the records, double the time.)

I have thus far disabled the Apache caching (both disk and memory), 
increased the memory limit for PHP from 8 to 16 MB, and , as previously 
mentioned, varied the size and number of record processed by the PHP 
script. None of these efforts, save the reduction in the amount of data 
requested, have had any effect in resolving Apache's use of CPU AFTER a 
page has loaded.

I very much appreciate your efforts thus far, they have been valuable. 
However, I would greatly appreciate any further insight you might have 
regarding this problem.

William Vaughn
Richard Carlton Consulting, Inc.