Mailing-List: contact test-dev-help@httpd.apache.org; run by ezmlm
Precedence: bulk
Reply-To: test-dev@httpd.apache.org
Message-ID: <304BDB72275BBB4DA590832B55A2029147277B@xsun04.ptp.hp.com>
From: "MATHIHALLI,MADHUSUDAN (HP-Cupertino,ex1)" <madhum@hp.com>
To: "'test-dev@httpd.apache.org'" <test-dev@httpd.apache.org>
Cc: dev@httpd.apache.org
Subject: RE: Regarding Apache 2.0.48 and specweb99
Date: Fri, 21 Nov 2003 14:05:31 -0500
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"


>-----Original Message-----
>From: gregames@apache.org [mailto:gregames@apache.org]
[SNIP]

>cgid should _never_ exit without something in the error log.  
>That makes it 
>sound like a core problem, i.e. ap_process_child_status() or a 
>signal handler is 
>fubar, in addition to whatever made the cgi daemon die.
>
>But if that is in fact happening, I would trace syscalls & 
>signals for the cgid 
>process.  (can't remember what the HPUX trace program is 
>called, but you want 
>something similar to truss/strace)

To give a background, here's what I did :

1. Use a large timeout and keepalive timeout, and 100 threads / process.
2. Use HTTP/1.0 as the SPECweb99 client seems to have some problem with
HTTP/1.1
   (not much work done there)
3. Start SPECweb99 run
   nabled all the different dynamic tests - DYNAMIC_CONTENT, DYNAMIC_POST,
DYNAMIC_CAD_GET, DYNAMIC_CGI_GET.
   - The run came back with a lot of "Can't connect" errors
     It's probably okay because some config was probably screwed up
4. The stipulated 20 min. warmup and the 20 min run happens
   The results are NOT posted even after 30 minutes
5. I get suspicious, and I try to do a simple GET to Apache
   - realized that Apache was hung. (telnet localhost 80... GET /foo  etc
stuff)
6. Attached gdb to each of the process - and found that a couple of
processes were processing do_post (in mod_specweb99) and NO cgid process
(YES - I backported Jeff's patch to restart cgid) !!

I tried attaching tusc to the cgi daemon - but since the daemon dies at a
random time, my log file was getting too full, and I had to just stop it. I
tried resetting the log a couple of times - but then I got diverted and
started thinking in a diffent angle : were the timeouts were too long, is
the system was running out of sockets etc.

I can reproduce the problem every single run of SPECweb99 (history: 2.0.43
ran just fine). I'll try to get the tusc for cgid when it dies - to see if
it helps.

BTW, one more thing I noticed : there's some problem with keeping the
sockets alive for a long time. The SPECweb99 client logs a error on close
socket (EBADF) when the keepalive times out.

-Madhu