Message-ID: <3352F451.A5827FB3@topsail.org>
Date: Mon, 14 Apr 1997 23:21:53 -0400
From: Chuck Murcko <chuck@topsail.org>
Organization: The Topsail Group
MIME-Version: 1.0
To: new-httpd@apache.org
Subject: Re: [STATUS] Sun Apr 13 19:08:09 PDT 1997
References: <Pine.NEB.3.96.970415024359.2937B-100000@localhost.imdb.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: new-httpd-owner@apache.org
Precedence: bulk
Reply-To: new-httpd@apache.org

Just ran some stress tests. I *am* still getting some SIGSEGVs, at a
very low rate, about the rate Rob's seeing with his. The core dumps I'm
getting are still untraceable, though.

I wonder if my gdb is whacked?

Plus, we've got 2 PRs about core dumps now. One from FreeBSD 2.2, and
one from Linux.

Marc posted a patch to allow core dumps on FreeBSD. Did you guys see
that?

Even at 1/5000, it's a problem if you're running millions of hits/day.

I think it's time to fire this up single threaded in the debugger and
beat it to death. Crude, but effective.

Rob Hartill wrote:
> 
> On Mon, 14 Apr 1997, Randy Terbush wrote:
> 
> > > Randy, are you still seeing these? I've gotten no core dumps in about a
> > > week of heavy testing since all of Roy's patches went in, and I fixed
> > > some other things in the proxy.
> >
> > No I am not seeing them. However, I _am_ seeing the following:
> > pid 9556 (httpd), uid 50002: exited on signal 6
> > pid 9597 (httpd), uid 50002: exited on signal 6
> > pid 8616 (httpd), uid 50002: exited on signal 6
> > pid 9816 (httpd), uid 50002: exited on signal 6
> > pid 9644 (httpd), uid 50002: exited on signal 6
> > pid 9645 (httpd), uid 50002: exited on signal 6
> > pid 21720 (httpd), uid 50000: exited on signal 4
> > pid 16767 (httpd), uid 501: exited on signal 6
> > pid 7406 (httpd), uid 501: exited on signal 6
> > pid 22712 (httpd), uid 501: exited on signal 6
> > pid 8163 (httpd), uid 50054: exited on signal 6
> > pid 8165 (httpd), uid 50054: exited on signal 6
> > pid 8164 (httpd), uid 50054: exited on signal 6
> >
> > Bus errors. Lots of em. I also spent some time this past weekend
> > trying to get a coredump out of FreeBSD. I have verified that
> > RLIMIT_CORE is of sufficient size. Despite the fact that Ken's
> > patch now helps find where we might be putting the core, and the
> > directory is world writeable, no core...
> >
> > I'll be looking at this more through the week.
> 
> I have "lots" of these too under FreeBSD 2.2
> No sign of any core files here either. I thought mine might
> be due to non-standard stuff I run.
> 
> Instead (or in addition to) saying *where* the core was meant to
> get dumped, I'd have more success in tracking this down if the
> offending URL was logged.. unless it turns out to be URL independent
> and triggered by something else (e.g. User-Agent).
> 
> Of course I could hack my own httpd to log the info I need, and
> maybe I will  :-). My SIGSEGV/hit ratio is ~ 1/5000 so it's not a
> major problem so far.
> 
> --
> Rob Hartill                              Internet Movie Database (Ltd)
> http://us.imdb.com/tour   .. a site for sore eyes.

-- 
chuck
Chuck Murcko
The Topsail Group, West Chester PA USA
chuck@topsail.org