Hi there,
I just wanted to report the same problem, apparently the situation is
improved in 3.2.2beta 2 .
What I had is that under 3.1 and 3.2.1 when stress loaded with 50 concurrent
users making request each second Tomcat leaked memory.
It would easily go to OutOfMemoryError, although I tried to help him by
adding a servlet that fired the GC.
So clearly it was a memory leak somewhere.
I could have blamed my application for that but I did the same with other
containers.
Now the memory leak apparently has gone, but under the same stress Tomcat
generates some extraneous errors (the problem was there before):
"2001-03-30 12:32:38 - Ctx( /pictures ): IOException in: R( /pictures +
/ImgCache
/4/400/4003[cr_400]40030015.jpg + null) Connection reset by peer: socket
write e
rror"
Is there a way to report these errors with some extra debug info ?
Under no stress, the problem does not apear.
Also the CPU usage for tomcat (800Mhz/256MB RAM)under 35 requests/second
goes to 50-60%,
while for resin 1.2.3 it bearly touches 15%.
So you're on the good path, but you still have a long way to go.
Hope this helps.
Costin
-----Original Message-----
From: Marc Saegesser [mailto:[EMAIL PROTECTED]]
Sent: Friday, March 30, 2001 11:29 AM
To: [EMAIL PROTECTED]
Subject: RE: Stress Test problems with 3.2.1
Could you please try this test with Tomcat 3.2.2 beta 2? I believe this
problem has been fixed in this release.
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]
> Sent: Friday, March 30, 2001 1:08 PM
> To: [EMAIL PROTECTED]
> Subject: Stress Test problems with 3.2.1
>
>
> Hello,
> I have tc 3.2.1 installed on w2k server with iis5, sun hotspot 2.0 server
> mode and jdk 1.3. I have been running this implementation
> successfully for
> the past several months with no problems. I decided to perform a stress
> test to see how tc would handle it.
>
> 1. First test. 100 concurrent connections, each submitting a request to
> /examples/jsp/forward.jsp every 10 seconds. This equated to 11
> requests per
> second. I ran this test for 24 hours. Results were perfect, no errors.
>
> 2. Second test. same test as above with 200 concurrent connections. This
> equated to 18 requests per second. Ran for 5 minutes. This time
> I received
> errors on 1% of the requests -- "500 - Internal Server Error". I checked
> the Tomcat log and I am getting NullPointerExceptions from the jsp code
> followed by an ArrayIndexOutOfBoundsException at
> tomcat.util.SimplePool.put(SimplePool.java:112). The redirect
> dll reports a
> few problems of not being able to handle the tc response. I suspect tc
> could not deliver a response because of the internal errors. Not sure.
>
> I go back to test 1 and everything is back to perfect. Go to test 2 again
> and same errors. As I lowered the number of requests per second, it gets
> better. It did not matter if I had 300 concurrent connections as long as
> the request per second was around 11, everything worked fine. Does anyone
> know what's up here?
>
> Thanks
> Mike
> [EMAIL PROTECTED]
>
