A lot of "segmentation fault", Ubuntu Hardy, apache with mod_perl

[Igor Chudov <ic...@gmail.com>]

I just moved my website algebra.com, from my old trusty Fedora 3 server, to
my brand new Ubuntu Hardy server. The Hardy server is fully patched.

In the logs, I see a lot of error messages saying

segmentation fault

in the error.log file. No other details like pids, etc are provided.

~==>apache2 -v
Server version: Apache/2.2.8 (Ubuntu)
Server built:   Jul 10 2009 18:41:09

I have no way to understand where these messages are from, how to debug the
fault, etc.

Any ideas? I have used linux for 14 years and would like to troubleshoot
this problem.,

Re: A lot of "segmentation fault", Ubuntu Hardy, apache with mod_perl

[Adam Prime <ad...@utoronto.ca>]

Igor Chudov wrote:
> 
> 
> On Sun, Aug 9, 2009 at 7:08 AM, Nick Wellnhofer <wellnhofer@aevum.de
> <ma...@aevum.de>> wrote:
> 
>     Igor Chudov wrote:
>     > In the logs, I see a lot of error messages saying: segmentation fault
>     Are you using the worker MPM? If yes, try the prefork MPM instead.
> 
> 
> Nick, thank you, per your advice, I did try using prefork MPM.
> 
> Unfortunately, the end result was still the same, a lot of segmentation
> faults after a while.
> 
> The symptom is always the same, it runs great for a while (like maybe
> half an hour) and then begins to give segmentation faults.
> However, no matter what pages of my site I visit at this time, they all
> display correctly.
> 
> I also ran apache2 with -X flag under GDB. I disabled SIGPIPE handling.
> 
> Again, it ran for an hour and crashed. However, when I tried to say
> "where" in GDB, GDB told me that the process no longer exists.
> 

Have you read this section of the docmentation on perl.apache.org?

http://perl.apache.org/docs/2.0/devel/debug/c.html

particularly the stuff about generating core dumps and analyzing them

Adam

Re: A lot of "segmentation fault", Ubuntu Hardy, apache with mod_perl

[Igor Chudov <ic...@gmail.com>]

On Sun, Aug 9, 2009 at 7:08 AM, Nick Wellnhofer <we...@aevum.de> wrote:

> Igor Chudov wrote:
> > In the logs, I see a lot of error messages saying: segmentation fault
> Are you using the worker MPM? If yes, try the prefork MPM instead.
>

Nick, thank you, per your advice, I did try using prefork MPM.

Unfortunately, the end result was still the same, a lot of segmentation
faults after a while.

The symptom is always the same, it runs great for a while (like maybe half
an hour) and then begins to give segmentation faults.
However, no matter what pages of my site I visit at this time, they all
display correctly.

I also ran apache2 with -X flag under GDB. I disabled SIGPIPE handling.

Again, it ran for an hour and crashed. However, when I tried to say "where"
in GDB, GDB told me that the process no longer exists.

i