segfault hell

[Bill Moseley <mo...@hank.org>]

I'm running current Apache and mod_perl 1.24.  Solaris 2.6. 2G RAM.

PerlFreshRestart is off.

I've had these segfaults for months.  I only get a few a day out of say
20,000 mod_perl requests.  So the children come and go before I can get
strace or gdb on them.  

They often happen in groups -- as if something funky happens to the OS and
a few segfault all at once.  It's not the URI that's triggering the
segfault, and I've never been able to trigger it myself.  It can happend
after just a few requests to a httpd child, or after many requests to a child.

Any other ideas on how to catch this in action without generating a huge
log file?  Set MaxRequestsPerChild huge and attach gdb to every child?  

Undefined subroutine &Apache::RegistryNG->handler::handler called at
/usr/local/lib/perl5/site_perl/5.005/sun4-solaris/Apache.pm line 184.

Then that child may server up a few more static files, but next time
there's a mod_perl request:

Can't locate object method "handler" via package "handler" at
/usr/local/lib/perl5/site_perl/5.005/sun4-solaris/Apache.pm line 184.

I am using a $SIG{__DIE__} handler in my Registry script, which just writes
to a log file and calls (the overridden) exit.



Bill Moseley
mailto:moseley@hank.org