http://code.google.com/p/modwsgi/wiki/ConfigurationDirectives#WSGIDaemonProcess

2012/6/14 Michele Comitini <michele.comit...@gmail.com>:
> I would still lower Threads to 5 and use the WSGIDaemonProcess
> directive with only processes.
>
> WSGIDaemonProcess processes=x
>
> x depends on how much memory you have.  since web2py *is* fast you do
> not need higher than x = 4*n cpus.
> with processes dal db connection can be with poolsize=1.
>
>
> mic
>
>
> 2012/6/14 Massimo Di Pierro <massimo.dipie...@gmail.com>:
>> I am switching to nginx as soon as I move the server. Anyway for now I
>> disabled apps that are not used (should not be used) and changed the apache
>> conf to
>>
>> Timeout 35
>> KeepAlive On
>> MaxKeepAliveRequests 15
>> KeepAliveTimeout 2
>> <IfModule mpm_worker_module>
>>     StartServers          5
>>     MaxClients           10
>>     MinSpareThreads      15
>>     MaxSpareThreads      15
>>     ThreadsPerChild      10
>>     MaxRequestsPerChild  50
>> </IfModule>
>>
>> and the memory exhaustion problem seems to be gone away. At least for now.
>>
>> massimo
>>
>> On Thursday, 14 June 2012 05:10:43 UTC-5, mcm wrote:
>>>
>>> Massimo,
>>>
>>> do you have any good reason to run apache?  if not switch to nginx.
>>>
>>> mic
>>>
>>>
>>> 2012/6/14 Michele Comitini <michele.comit...@gmail.com>:
>>> > Correction:
>>> >> mpm_worker -> spawn a number of different long running processes to
>>> >> manage requests
>>> > mpm_worker -> spawn a number of different long running *threads* to
>>> > manage requests
>>> >
>>> >
>>> > 2012/6/14 Michele Comitini <michele.comit...@gmail.com>:
>>> >> mpm_prefork -> fork a number of different long running processes to
>>> >> manage requests
>>> >> mpm_worker -> spawn a number of different long running processes to
>>> >> manage requests
>>> >>
>>> >> wsgi can fork and thread python interpreters so the overall status of
>>> >> forking and threading varies a lot because of this.
>>> >> An important thing to remember is that if the python interpreter dies
>>> >> all threads in the same thread group die.
>>> >> If there is a pure threaded only configuration the whole apache server
>>> >> can die if python dies.
>>> >>
>>> >> Switching to prefork could make the site more available, but does hide
>>> >> the real problem, i.e. the python interpreter dies for some reason.
>>> >>
>>> >> N.B. php can work reliably only on prefork configuration, because it
>>> >> is not thread safe.
>>> >>
>>> >> mic
>>> >>
>>> >>
>>> >> 2012/6/14 Massimo Di Pierro <massimo.dipie...@gmail.com>:
>>> >>> some follow up. One thing that happened when the problem started is
>>> >>> that I
>>> >>> was experimenting with the apache settings and turned off keepalive.
>>> >>>
>>> >>> I do not fully understand the implications of apache settings
>>> >>> (mpm_prefork_module vs mpm_worker_module) and their parameters but
>>> >>> since
>>> >>> Python get slower the more cores are available for a fixed number of
>>> >>> threads, there is an optimum set of parameters depending on the
>>> >>> architecture.
>>> >>>
>>> >>> I will continue some experimentation.
>>> >>>
>>> >>> Massimo
>>> >>>
>>> >>>
>>> >>> On Wednesday, 13 June 2012 17:47:55 UTC-5, Massimo Di Pierro wrote:
>>> >>>>
>>> >>>> The fact is I did not have that problem before. The number of users
>>> >>>> has
>>> >>>> not increased much. Either a recent commit changed something (if this
>>> >>>> is a
>>> >>>> web2py problem) or it is a cache issue (I have many apps and some
>>> >>>> cache too
>>> >>>> much stuff in ram).
>>> >>>>
>>> >>>> massimo
>>> >>>>
>>> >>>> On Wednesday, 13 June 2012 17:38:26 UTC-5, Bruce Wade wrote:
>>> >>>>>
>>> >>>>> This seems like the same problems I am having with
>>> >>>>> new.youadworld.com all
>>> >>>>> the resources are getting sucked out of 5 servers. I am starting to
>>> >>>>> think I
>>> >>>>> will have to start porting some of the sites features to a different
>>> >>>>> technology.
>>> >>>>>
>>> >>>>> This is a bad situation as I have been developing
>>> >>>>> www.fittraineronline.com also with web2py however I cannot run into
>>> >>>>> the same
>>> >>>>> situation with that site, especially when I will have a mobile web
>>> >>>>> app also
>>> >>>>> with potentially millions of members.
>>> >>>>>
>>> >>>>> I am going to continue digging through the core and try to optimize
>>> >>>>> this
>>> >>>>> as the developers productivity is great. However spending around
>>> >>>>> $1200 a
>>> >>>>> month just to run a site that can't handle 40,000 members isn't
>>> >>>>> good.
>>> >>>>>
>>> >>>>> On Wed, Jun 13, 2012 at 3:31 PM, Massimo Di Pierro
>>> >>>>> <massimo.dipie...@gmail.com> wrote:
>>> >>>>>>
>>> >>>>>> I do not know. Everything worked fine for long time until one week
>>> >>>>>> ago I
>>> >>>>>> started having problems with the server. The fact is nothing
>>> >>>>>> changed in
>>> >>>>>> web2py to be causing this. I am still debugging but not much info
>>> >>>>>> to look
>>> >>>>>> at. I have lots of apps running on the same server and I am
>>> >>>>>> planning to do a
>>> >>>>>> major cleanup so we can have mirrors.
>>> >>>>>>
>>> >>>>>> Massimo
>>> >>>>>>
>>> >>>>>> On Wednesday, 13 June 2012 16:13:25 UTC-5, JavierQQ wrote:
>>> >>>>>>>
>>> >>>>>>>
>>> >>>>>>> On Wed, Jun 13, 2012 at 4:08 PM, davidjensen  wrote:
>>> >>>>>>>>
>>> >>>>>>>> cannot establish connection to www.web2py.com
>>> >>>>>>>> 2012-06-13T17:11
>>> >>>>>>>
>>> >>>>>>>
>>> >>>>>>> Me too :(
>>> >>>>>>> why is this happening constantly?
>>> >>>>>>>
>>> >>>>>>> Javier
>>> >>>>>
>>> >>>>>
>>> >>>>>
>>> >>>>>
>>> >>>>> --
>>> >>>>> --
>>> >>>>> Regards,
>>> >>>>> Bruce Wade
>>> >>>>> http://ca.linkedin.com/in/brucelwade
>>> >>>>> http://www.wadecybertech.com
>>> >>>>> http://www.fittraineronline.com - Fitness Personal Trainers Online
>>> >>>>> http://www.warplydesigned.com
>>> >>>>>
>>> >>>

Reply via email to