Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:13052
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
To: internals@lists.php.net
Organization: Last.fm
Date: Tue, 28 Sep 2004 18:03:34 +0100
User-Agent: Pan/0.14.2.91 (As She Crawled Across the Table (Debian GNU/Linux))
Message-ID: <pan.2004.09.28.17.03.02.555458@last.fm>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 8bit
Subject: Really odd PHP problem (redux)
From: russ@last.fm (Russ Garrett)

(The original thread is here:
http://www.zend.com/lists/php-dev/200409/msg00073.html)

OK, so quite a lot has changed since my last post. The web server is now
running Debian 3.1 as opposed to 3.0, and I've been horrendously busy so I
haven't had time to investigate this any further.

However, our trial of ZPS is about to expire, and although it's a great
bit of software, to be completely honest we don't really have the the
money to buy it currently. So I switched back to mmcache to see if the
problem had gone away.

It hasn't, and we're still getting the problem detailed in my initial
post: in short, every so often a process will eat 25% of the RAM and hang.
After a while, so many of these processes accumulate that they bring the
server down.

I have managed to catch one of these processes and analyze it, a log of
which is available here:

http://static.last.fm/phpbug/1.txt

I did try APC again as an alternative to mmcache, however that still
doesn't work for me, I managed to collect some more info here, however I
don't have time to pursue this bug much further at the moment (I swear I
have the worst luck ever):

http://static.last.fm/phpbug/apcbug.txt

Is it still likely that mmcache is the cause of this? If so, any ideas on
the APC bug?

Thanks in advance,

Russ Garrett
russ@last.fm