Re: Summing up the Icebreak discussion

[Nathan Andelin <nandelin@xxxxxxxxx>]

Kevin,

In the case that I referred to earlier, we set ThreadsPerChild to 3000, and over 
a period of time the CGI jobs grow.  We use a dispatcher architecture similar to 
yours.

-Nathan


----- Original Message ----
From: Kevin Turner <kevin.turner@xxxxxxxxxxxxxxx>
To: Web Enabling the AS400 / iSeries <web400@xxxxxxxxxxxx>
Sent: Fri, January 7, 2011 1:35:16 AM
Subject: Re: [WEB400] Summing up the Icebreak discussion

Hi Nathan

Some installations run with the despatcher (the data queue isolating the CGI 
stuff from the Apache thread) and some do not.  In this case they do not - the 
programs are all running in the thread. They are going to switch to use the 
despatcher this month to see if it improves matters.

The odd thing is that it will behave for a week or so and then suddenly start to 
issue thread failure messages to the QSYSOPR message queue. Sometimes that just 
stops and other times it escalates until the server just crashes completely and 
has to be restarted. I am sorry I haven't got the job log messages to hand, but 
they are occurring in the bowels of the Apache software somewhere. It all sounds 
like some sort of memory leak issue, but we haven't isolated one in the 
application yet.  IBM were not much help either.

"1-2 thousand CGI server jobs".  Wow!  No, nothing of that magnitude.  They have 
max out at just over a 1000 concurrent sessions during the day - about 3-400 
call centre agents I guess, each with an average of 2 sessions (some more).  
Those sessions are usually served by a 100 or so CGI threads - could be more but 
certainly not a 1-2 thousand. That is an enormous number - are they running 
persistent CGI?

Cheers
Kevin