Mailing-List: contact tomcat-user-help@jakarta.apache.org; run by ezmlm
Precedence: bulk
Reply-To: tomcat-user@jakarta.apache.org
Sender: andym
Message-ID: <3AE3C3F4.9E496137@hotcopper.com.au>
Date: Mon, 23 Apr 2001 15:56:04 +1000
From: Andrew McDonald <andy@hotcopper.com.au>
Organization: HotCopper Australia
MIME-Version: 1.0
To: tomcat-user@jakarta.apache.org
Subject: 2 Critical Problems with Tomcat 3.2.x
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Hi there,

First I'd like to thank the developers for a great product.  Tomcat is a
nice piece of software.  And like all great OSS, very easy to raise a
purchase order for ;)  Thanks guys.
Unfortunately I have run into a couple of serious show-stoppers when
deploying TC in production.
A bit of background:
We have been running various incarnations of JServ and Tomcat over the
past 12 months or so.
We have 3 Application Servers: Two Compaq Deskpro PII/450s and one
PIII/500 with 128MB Ram each. They are running Linux, kernel is 2.2.19,
glibc-2.1.3.  We are currently running various incarnations of TC3.2.x
We run a Stock Trading Forum, which is made up of Servlets and JSPs.
You can see it at http://www.hotcopper.com.au if you're interested at
all.  We are one of the most hit websites in Australia according to
www.top100.com.au.

We have tested Tomcat with the Solaris 1.3 JVM, IBM 1.3 JVM and
Blackdown 1.3 JVM (FCS).  I'm happiest with Blackdown.  It *seems* to
have the best performance.

On to the problems:
1.  We dubbed this one "The Apache White Screen" problem.  We connected
Tomcat to various versions of Apache <=1.14 in the standard way
(mod_jk.so)  Symptom:  People using Internet Explorer to browse our site
will intermittently get a white/ blank screen instead of results from
our application.  When this happens, we believe the request doesn't even
enter the java code, so it may be a problem with the connector.  I seem
to recall (sorry this was about 8 months ago) Apache returning a 304 or
a 302.  I haven't re-tested this in a while, I was just wondering
whether people were aware of this problem in the past and had already
looked into it and possibly fixed it.
Our "solution" was to run Tomcat standalone, which seemed to solve this
issue.

2.  We dubbed this one "The 132 Thread" Problem, even though it doesn't
always happen with 132 threads; sometimes 128, 130, or if using the
Blackdown JVM, 134.  (wierd).   We use NATIVE threads.  Symptom:  Tomcat
stops responding to http requests after a certain number (~132) of
threads is reached.  When this happens it will either deadlock or
segfault.  During this time the box is fine, and to solve it temporarily
I just need to kill all the java processes and restart Tomcat.  This
problem occurs regardless of JVM and the version of the Linux Kernel
doesn't seem to make a difference.  It happens once or twice a week.

Assistance with either of these problems would be greatly appreciated.
Number 2 is the really important one but information regarding number 1
would be great as well.

Of course if you need more information please get back to me.

Oh, and btw, yes I realise that running Compaq desktops in production
isn't the best idea ;)  That will be attended to.

Thanks in advance

Cheers

--
Andrew McDonald                "We cannot reason ourselves out of our basic
Systems Administration         irrationality. All we can do is learn the art
HOTcopper Australia              of being irrational in a reasonable way."
http://www.hotcopper.com.au                      - Huxley