On Sun, Jul 17, 2011 at 12:57 PM, Rafael Hasson <rafaelhas...@gmail.com>wrote:
> Hello Jeff, > > Seems, that the issue was with PHP PECL APC Cache > > I increased the kern.ipc.shmma on FreeBSD, until now no crash. > Will wait to see, if crash again I will do a backtrace with GDB! > thanks for updating the thread BTW, "crash" and "hang" are very different problem symptoms. You encountered a "hang," not a "crash." (Sometimes a better description of a symptom gets a faster resolution.) > > > ---------------------------------------------------------------------------------------------------------------- > Re: > [users@httpd] Apache crashing - *lockf* > status<http://markmail.org/message/wbvuj7gyxqhmdqax>[image: > permalink] <http://markmail.org/message/wbvuj7gyxqhmdqax> Jeff Trawick ( > traw...@gmail.com)Jul 12, 2011 5:28:17 pmorg.apache.httpd.users > > > -------------------------------------------------------------------------------------------------------------------------------------------- > > On Fri, Jul 8, 2011 at 5:34 PM, Rafael Hasson <rafa...@gmail.com> wrote: > > *Hello all! * > > *I've a webserver running PHP 5.2.17, Apache 2.2.17 and MySQL 5.5.13. > Tables' storage engine are MyISAM. I'm having a problem with all httpd > processes locking during peak hours. Today while investigating the problem, > I could replicate the it (all httpd processes locking) while running a 2.2 > seconds select statement. Does someone have a clue why a table locked for > 2.2seconds is making all httpd processes goes to lockf state? On my > understanding only MySQL should be on hold until statement completes, not > httpd. * > > *After httpd goes to lockf, it never returns, I have to kill it and start > again. This is how top lock then http://pastebin.com/uG7ZjD7A* > > -------------------------------------------------------------------------------------------------------------------------------------------- > > > Get backtraces of httpd processes to see where httpd is blocked. > > prefork with multiple listening sockets? Maybe the subset of httpd > processes in *lockf* state are just waiting for an incoming connection? > Backtraces will show. > > > > > -- Born in Roswell... married an alien...
