On Thu, Nov 25, 1999, Johan Ekenberg wrote:
> Apache dies apparantly at random occasions. The root-thread dies and all
> it's children slowly timeout one by one. There is *no* sign at all in the
> logs, no error message, no common circumstance either in apache-logs or
> systemlogs. Not even a "caught ???TERM, shutting down", it just stops
> logging and dies. The coredump shows it's a segfault. Everything else on the
> machine works very well, and so does Apache when it's running. One thing I
> have discovered though is that restarting Apache (either via "apachectl
> stop/start" or "apachectl gracful") appears to increase instability. The
> machine we are currently adding more customers to is much more instable
> since Apache is restarted frequently. I've tried compiling without
> mod_throttle, thinking that an error occured due to my fiddling with its
> source - no luck, Apache keeps crashing.
>
> Since Apache segfaults some 30-50 times each day, I had to write a small
> script which is run by cron each minute, restarting Apache each time it
> crashes. Not so good.
>
> I'm asking on the mod_ssl-list because the problem-report closest matching
> our problem is in the announcement of mod_ssl-2.4.8-1.3.9. This makes me
> think it might have to do with mod_ssl.
>
> SETUP:
> We're currently using mod_ssl 2.4.8 on 3 similar machines with this setup:
> - Linux Slackware 4.0, kernel 2.2.12 / 2.2.13ac2
> - Apache 1.3.9
> - mod_perl/1.21
> - PHP/4.0b3
> - mod_ssl/2.4.8
> - OpenSSL/0.9.4
> - mod_throttle-2.06
Just because you read that mod_ssl had segfault problems in the past, doesn't
mean your problem is related to mod_ssl, of course. Look at the large amount
of packages you use. They can be the source of your segfault with the same
possibility as mod_ssl+OpenSSL. Actually (because you say you're using >=
2.4.8) I think the segfault didn't happen inside mod_ssl. But, as the mod_ssl
FAQ clearly tells you, until you debug it with a debugger and at least find
out the code _location_ where the segfault occurs, _nobody_ can really help
you.
> Each machine has approx. 1100 VirtualHosts
> mod_throttle is patched to accept up to 2500 VirtualHosts,
> apache/src/httpd.h has HARD_SERVER_LIMIT raised to 1024. All limits in
> httpd.conf are set high, so are the ulimits f�r the process. I don't beleive
> there's a problem with file-descriptors or max-processes, since that tends
> to cause an error-message in the system-logs (?).
>
> I'd really appreciate some help on this, I don't know what to do next.
As I said, no one can help you just with this amount of details. You've to
attach a debugger to your stuff or at least post-inspect the core files and
find out more. Please read the mod_ssl FAQ for a few hints about this.
Ralf S. Engelschall
[EMAIL PROTECTED]
www.engelschall.com
______________________________________________________________________
Apache Interface to OpenSSL (mod_ssl) www.modssl.org
User Support Mailing List [EMAIL PROTECTED]
Automated List Manager [EMAIL PROTECTED]
