httpd-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chris <>
Subject Re: [users@httpd] Apache 1.3 vs Apache 2
Date Sat, 10 Dec 2005 11:15:45 GMT
On 07/12/05, Michael Jeung <> wrote:
> Good morning all,
> Recently, we've migrated one of our production servers from a Linux
> box to a FreeBSD box.  We're running Apache 1.3.34, with a PHP/MySQL
> web application.  Listed below, you can find how apache has been
> compiled.
> Yesterday afternoon, we put this new server into production and it
> seemed to be behaving relatively OK, with system-load of 4-5.  We
> thought everything was going well and that this issue was wrapped
> up.  Serves us right - today we got into the office and found that
> loads on our server had spiked to 150.
> Before we pulled the server from production, we grabbed a few
> snapshots of what the system was doing.  After taking a look at
> these, I haven't been able to make much head-way.  If someone with
> more experience in this matter could take a look, I would greatly
> appreciate it.
> Ultimately, my goal here is to get Apache to behave.  Any solution
> that will allow Apache to run without killing the server is
> acceptable -- including upgrading to Apache 2.  (Does Apache2
> outperform Apache 1.3?)

Hi I recently recovered from a similiar problem and mysql was the culprit,
it looked innocent sitting there using 0% cpu but when I ran mytop I seen
dozens of threads just sitting there so it had somehow locked up causing a
chain reaction to apache.  Of course restarting apache and leaving mysql
alone would recover the situation for a temporary period because it killed
all these open connections to mysql unclogging it.  What I did to resolve
the problem was (a) disable keepalive on apache, which helped a great deal,
(b) changed the threading on mysql I also stopped using WITH_PROC_SCOPE_PTH
and disabled kern.threads.thr_scope_sys, I still have mysqld statically
compiled tho.  (c) I did change a few things in my.cnf like reduce thread
caching and timeout values but I think (a) and (b) were what solved it, I am
not 100% confident yet but I am close to 48 hours without a hitch and was
having to restart apache numerous times a day.


View raw message