Mailing-List: contact dev-help@httpd.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@httpd.apache.org
Message-ID: <3F65A60C.3030105@attglobal.net>
Date: Mon, 15 Sep 2003 07:44:12 -0400
From: Jeff Trawick <trawick@attglobal.net>
Reply-To: trawick@attglobal.net
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US;
 rv:1.4) Gecko/20030624
MIME-Version: 1.0
To: Gagan Puri <gaganpuri@infosys.com>
CC: dev@httpd.apache.org
Subject: Re: Strange behaviour with IBM IHS 2.0.42.2 on HP 11i
References: <2B721C6525F0D411B1E900B0D0226BDD0338AD1D@mohmsg01.ad.infosys.com>
In-Reply-To: 
 <2B721C6525F0D411B1E900B0D0226BDD0338AD1D@mohmsg01.ad.infosys.com>
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

Gagan Puri wrote:

> Long time back there was issue related HP Apache 2.0.43. in which a method calls another 
 >method & hangs the execution. In resolution to that there was a 
suggestion that this
 >could be a known issue with hpapache2 (JAGae57148) where it gets a 
thread stack overflow.
 >It was also told that this bug was fixed on HP Apache 2.0.45 which 
indeed was fixed on
 >HP Apache 2.0.45. This issue was faced on HP 11i m/c.
> 
> I am facing the similar issue with IBM IHS 2.0.42.2 on HP 11i. The IBM IHS 2.0.42.2 webserver 
 >is core apache 2.0.46 based. The behaviour is same as what was 
mentioned with HP Apache 2.0.43.

Have you tried real Apache as downloaded from httpd.apache.org?  If 
there is a suspected stack overflow problem I imagine you'll hit the 
same symptom with real Apache too, and we can figure out what Apache 
code to change to bump up the stack size and see if that resolves the 
problem.

 >The behaviour is as below.

It sounds like it could be related to stack overflow, but I don't know 
for sure.

> The same application works fine with core apache 2.0.46, and with many other webservers. 

Pure Apache 2.0.46 as downloaded from httpd.apache.org, or HP's Apache?

> Can this stack overflow also occur on IBM IHS 2.0.42.2(apache 2.0.46 based). Is this 
 >entirely web server specific or something can be done specific to my 
application to get rid of this issue.

Apache and APR (and IHS as well) have no provision to change the thread 
stack size from the default*.  I have heard from another vendor that 
HP's Apache has some provision to change the thread stack size from the 
default.  I don't know what that is.

*on AIX there is a system-provided environment variable to adjust the 
stack size for threads that use the default stack size, but I've never 
found anything like that for HP-UX

If indeed you're hitting a thread stack size limit, I think Apache 
and/or APR need to support the user modifying the thread stack size, but 
of course you could potentially change your application to reduce the 
stack usage and instead utilize the heap more.

Here is how I would suggest proceeding:

1) try it with real Apache, verify that same type of problem occurs, let 
us know what happens
2) assuming that you hit the same type of problem, apply a patch (to be 
determined) to bump up the thread stack size and see if that resolves it
3) assuming that it resolves it, we Apache httpd-ers need to discuss how 
to support such an application

If a problem is to be discussed on dev@httpd, it should be described in 
terms of Apache httpd anyway (not HP's Apache-based server, not IBM's 
Apache-based IHS, not some other patched version of Apache httpd).