[ic] IC + RedHat 7.x + Apache 1.3 = Segmentation fault?

Ed LaFrance interchange-users@interchange.redhat.com
Tue May 21 15:03:00 2002


At 01:52 PM 05/21/2002 -0400, you wrote:
>Quoting Matthew T. Snyder (snydermt@ariontech.com):
> > Ed LaFrance wrote:
> >
> > > As a result of my consulting business I have access to may servers, and
> > > I've noticed a condition that appears in varying degrees on just
> > > about any
> > > machine that is running RH 7.x or a similar flavor of *nix,
> > > Apache 1.3x and
> > > Just about any Interchange version. The apache error_log on said machine
> > > contains entries like this:
> > >
> > > [Sun Sep 23 09:45:22 2001] [notice] child pid 6913 exit signal
> > > Segmentation
> > > fault (11)
> > > [Thu Sep 27 13:58:00 2001] [notice] child pid 7236 exit signal
> > > Segmentation
> > > fault (11)
> > >
> > > Frequency may be once a month, once a day or once every ten minutes - it
> > > seems to depend upon how may IC daemons are running on the server (all
> > > other things being equal, more daemons = more frequent segfaults).
> > >
> > > There is no discernable performance impact or other undesirable
> > > manifestation that I can detect, other than the error_log being
> > > larger than
> > > it would otherwise. I tried to compare the error_log entries to those in
> > > access_log and the system messages log, but have not found any pattern of
> > > correspondence.
> > >
> > > Because the properties of sampled severs do vary, the only thing
> > > I can say
> > > with reasonable certainty is that I've only seen this on machines running
> > > Interchange - spot checks of a few that don't show no such
> > > error_log entries.
> > >
> > > Can anyone comment on this. At first glance, lack of physical RAM was
> > > highly suspect, but so far tests with greatly increasing RAM and greatly
> > > reducing apache MaxClients has only freed up memory - frequency of the
> > > annoying error_log entries has not dimished. I have not yet
> > > played with IC
> > > server settings, but FWIW, in most cases it is running n low traffic mode
> > > with the default settings for the same. IC version varies widely.
> > >
> >
> > Granted, the scenario I'm running here isn't under near the load as some of
> > your servers, but with RedHat7.2, IC 4.8.3, 512MB ram, and low traffic
> > settings, I'm not seeing any segmentation faults.  Should point out this is
> > not much more than a dev box at the moment, while I learn more about
> > interchange and tinker with it, so it probably only sees 2000 hits/day max.
> > That could be part of the issue.  The errors in the apache log for me 
> mostly
> > show some poor joker trying to take advantage of a vulnerability in another
> > web server (not naming any names, of course:))
> >
>
>This is almost undoubtedly the Perl signal problem. If you search for the
>long-running thread on PGP and PayFlow problems with completing transactions,
>the solution will boil down to running in PreFork mode with MaxServers set
>to 0 to disable all signals.
>
>--
>Mike Heins

Mike - thanks for this. I'll give it a try with the 4.8.x daemons on that 
server - I'll guess I'll just have to cope with the 4.6.x and Minivend 3.15 
(yep - still running it!) processes, since it appears that PreFork was not 
implemented until 4.7, unless there is some undocumented technique out there...

- Ed L.


===============================================================
New Media E.M.S.               Software Solutions for Business
463 Main St., Suite D          eCommerce | Consulting | Hosting
Placerville, CA  95667         edl@newmediaems.com
(530) 622-9421                 http://www.newmediaems.com
(866) 519-4680 Toll-Free       (530) 622-9426 Fax
===============================================================