Bug#215836: apache uses lot's of CPU and Memory

[Daniel Pocock <daniel@pocock.com.au>]

Thanks for the suggestions Matt.

The problem reoccurred only once today.   Running the debugging version 
of Apache.

I attached to the process in question with gdb.  gdb loaded the symbol 
info from /usr/sbin/apache.dbg   However, the backtrace command would 
still only produce the same output as below - so I still don't know 
which part of Apache is causing the problem.

I tried the step command in gdb, and found that it never left the 
function it was in.  I believe the function in question is part of the 
underlying libraries.

Is there anything else I can do to get more information out of gdb or to 
get Apache to log what it's doing to a logfile?

Matt Zimmerman wrote:

> On Wed, Oct 15, 2003 at 07:45:42PM +0100, Daniel Pocock wrote:
>
>  
>
> > I still have the problem with mod_jk disabled completely and all JkMount 
> > directives removed.
> >
> > Is there some way I can debug the process that is playing up?  I 
> > attached to it with gdb /usr/sbin/apache 18038 and each time I stop it I 
> > get:
> >
> > 0x40205ce4 in _IO_file_underflow () from /lib/libc.so.6
> > (gdb) cont
> > Continuing.
> >
> > Program received signal SIGINT, Interrupt.
> > [Switching to Thread 16384 (LWP 18038)]
> > 0x40205ce4 in _IO_file_underflow () from /lib/libc.so.6
> > (gdb)
> >
> > When I continue the process, the problem is back within 15 seconds or 
> > so.  Is there a symbol table I can use for debugging this binary, or 
> > will I need to recompile?
> >
> > (gdb) bt
> > #0  0x40205ce4 in _IO_file_underflow () from /lib/libc.so.6
> > #1  0x40205f65 in _IO_file_underflow () from /lib/libc.so.6
> > #2  0x40207aad in _IO_default_uflow () from /lib/libc.so.6
> > #3  0x402078ee in __uflow () from /lib/libc.so.6
> > #4  0x401fc328 in _IO_getline_info () from /lib/libc.so.6
> > #5  0x401fc223 in _IO_getline () from /lib/libc.so.6
> > #6  0x40204b4b in fgets_unlocked () from /lib/libc.so.6
> > #7  0x486bd008 in ?? ()
> > #8  0x0fbb8fff in ?? ()
> > #9  0x0000000a in ?? ()
> > #10 0x00000001 in ?? ()
> > (gdb)
> >
> >
> > My httpd.conf has about 20 virtualhosts and almost everything else is 
> > the defaults.
> >
> > There is nothing unusual about any of the virtualhost definitions.  All 
> > my sites use php.  I also have squirrelmail, but the problem still 
> > occurs even when no one has been using squirrelmail.  Eg, I just killed 
> > all the processes now, restarted, and the problem was back within 15 
> > minutes.
> >    
>
> It's possible that it's a UML bug, an underlying library like glibc, or a
> bug in Apache which is only exposed by UML (it sometimes uncovers race
> conditions which are unlikely in other situations).  You'll need to
> experiment with it to narrow the list of possible causes and determine which
> is at fault.
>
>