El mié, 28-07-2010 a las 16:17 -0700, Graham Dumpleton escribió:
> 
> On Jul 28, 11:12 pm, Agustin Henze <[email protected]> wrote:
> > El mar, 27-07-2010 a las 16:03 -0700, Graham Dumpleton escribió:
> >
> > > Please better define 'apache proccesses an unlimited increase'.
> >
> > Apache processes start to grow without limits[0].
> 
> Okay, so the number of processes grows, rather than the amount of
> memory used by a process. That is what I wanted to clarify.
> 
> > The configuration of
> > the VH is:
> > WSGIDaemonProcess trac user=www-data group=www-data processes=1
> > threads=1 maximum-requests=1000 display-name=wsgi-trac
> 
> Do you have WSGIProcessGroup and WSGIApplicationGroup directives in
> your configuration as well? Post the more complete configuration.
> 
> As is, your graph suggests classic problem of running stuff in
> embedded mode with prefork MPM. That is, chance you aren't delegating
> Trac to daemon process group properly. See:
> 
>   http://blog.dscpl.com.au/2009/03/load-spikes-and-excessive-memory-usage.html
> 
> Even if daemon mode used, your use of a single process with single
> thread is a bad idea. This is because you can only handle one request
> at a time. If for some reason a request blocks or takes a really long
> time, then no other requests can be handled at the same time. What
> will happen then if using prefork MPM is that Apache will create more
> and more Apache server child processes to accept the pending requests,
> but they will all block waiting to be able to proxy to daemon process.
> Since for prefork MPM the default is 150 processes, you can see as
> many as that number of processes created as requests arrive.
> 
> As such, if any extensions are thread safe (Trac core is already),
> then you should run more than a single thread. If code is still
> blocking or taking a long time, those additional threads will still be
> able to handle requests. If those all threads get stuck in same way,
> then still can have same issue.
> 
> if you are not using PHP, you should also change to worker MPM.
> 
> BTW, what version of mod_wsgi are you using?
> 
> Graham

Sorry for my long lag, i don't had much time for writing the answer. The
short history was, install apache mpm worker with mod_wsgi, 4 proccess
and 10 threads. After that, it works greatly.
Thanks Graham :)


-- 
TiN

Attachment: signature.asc
Description: This is a digitally signed message part

Reply via email to