I have a dual processor xeon server, running Red Hat Linux 9, that I was trying to stress test. The system has a SCSI based RAID 5 disk array. It had great response times for what I was testing until I got 468 instances of the application being tested running. It is a web based application that uses a cgi to parse a web form and pass data to my process via a socket connection. The application under test runs continually and waits for the cgi to hand it data. The data is then used to generate a dynamic HTML page that gets sent to the user's browser where they fill in the form and pass the form to the cgi again. I have a browser replacement that I run as many instances as the number of processes under test, on a second dual processor xeon server. The browser replacement auto fills the forms and submits them at some controlled pace. The browser system was just coasting doing it's thing.
If I replaced the cgi with a perl script and use mod-perl then I was able to run about 700 instances of the application being tested before the performance problem reappeared. It seems like some kind of system resource or system table is getting full and when I add the one extra instance of my application a wall gets hit and the new process goes into a wait state or something like that. I do not think it is CPU related as just before the process "that broke the camel's back" gets launched the system is only running at about 50% on each processor. Additionally memory does not seem to be an issue either. The CPU and memory usage is still reasonable after the troubled process starts. What I was wondering is what system monitoring tools exist to allow me to view various system tables in real time to see if I can figure out what resource needs to be increased. Any words of wisdom as to ways to try to figure out what barrier I might be hitting? Thanks for any ideas! -- Jim Dickenson mailto:[EMAIL PROTECTED] -- redhat-list mailing list unsubscribe mailto:[EMAIL PROTECTED] https://www.redhat.com/mailman/listinfo/redhat-list
