Ok so your machine really behaves concurrently it could be that you stumbled in some threading problem, probably at db driver level, I am quite confident about rocket, so the next questions are ;-) do you use a database? which database?
mod_wsgi is not so bad it is difficult to tune IMHO and does not scale because of apache. uwsgi uses many different models. Using uwsgi a mix of event + prefork could give a big boost, but if you care about performance avoid apache. Try nginx or cherokee as http server then scgi. We using that type of configuration in production with at least 180K json requests a day and very complex business logic with high concurrency. scgi is not much tunable but is really simple: on a machine like yours will do good. I wrote a slice here for a standard preforking scgi setup: http://www.web2pyslices.com/slices/take_slice/143 mic 2011/9/28 Phyo Arkar <phyo.arkarl...@gmail.com>: > Server machine is 32 cores 64 GB RAM :) > > WSGI limitation is bad at concurrency right? I am thinking to move to more > Concurrent setup. Coz version 1 of this software gonna use Qooxdoo for UI > and there will be a lot more concurrency. > > What are the advantages of uwsgi can use uevent right? > > On Thu, Sep 29, 2011 at 12:40 AM, Michele Comitini > <michele.comit...@gmail.com> wrote: >> >> Can you try with a setup without rocket? I.e. scgi or uwsgi or fastcgi. >> Is the server machine multicore? >> >> mic >> >> Il giorno 28/set/2011 18:48, "Phyo Arkar" <phyo.arkarl...@gmail.com> ha >> scritto: >> > Happened once again >> > >> > THis time 1st json request went ok and another request failed like it >> > did before. >> > >> > On 9/28/11, Phyo Arkar <phyo.arkarl...@gmail.com> wrote: >> >> Yeah i love screen coz i can search back easily and interactively debug >> >> infos. >> >> >> >> Yes its doing concurrent , the problem happen again yesterday but just >> >> for once , and gone away again. >> >> >> >> Theres 2 Grid (jqGrid) in one page , they request 2 json response from >> >> that controller but from different function at Once. >> >> >> >> On 9/28/11, Michele Comitini <michele.comit...@gmail.com> wrote: >> >>> No way is not bad practice IMHO screen is a great tool. >> >>> So we can exclude the problem of stdout buffer full. >> >>> Seems something low level. >> >>> >> >>> Is the client doing concurrent queries? >> >>> >> >>> mic >> >>> >> >>> >> >>> >> >>> 2011/9/27 Phyo Arkar <phyo.arkarl...@gmail.com>: >> >>>> Web2py is ran inside screen. coz my app is not production-stable yet >> >>>> and want to debug. I print out some debug messages and can track back >> >>>> inside screen (i know i know , its bad practice :D) >> >>>> >> >>>> may be coz of that? >> >>>> >> >>>> On 9/27/11, Michele Comitini <michele.comit...@gmail.com> wrote: >> >>>>> I would begin to remove possible environmental causes... ;) >> >>>>> Other questions: >> >>>>> 4) Do you run web2py.py attached to a tty? >> >>>>> 5) Do you redirect standard output from web2py or leave it on the >> >>>>> tty? >> >>>>> >> >>>>> The questions are because I see some print statements that could >> >>>>> hang >> >>>>> IO on a thread if the web2py process is not properly detached from >> >>>>> the >> >>>>> terminal. >> >>>>> If that is the case use nohup and try to see what happens. >> >>>>> >> >>>>> wrong: >> >>>>> $ python web2py.py <options> & >> >>>>> >> >>>>> right: >> >>>>> $ nohup python web2py.py <options> & >> >>>>> >> >>>>> If nothing changes... well... other questions! >> >>>>> >> >>>>> >> >>>>> mic >> >>>>> >> >>>>> 2011/9/27 Phyo Arkar <phyo.arkarl...@gmail.com>: >> >>>>>> It cannot be reproduced .. >> >>>>>> >> >>>>>> On 9/27/11, Phyo Arkar <phyo.arkarl...@gmail.com> wrote: >> >>>>>>> its not the same warning notices , client recieves 0 byte and >> >>>>>>> timed >> >>>>>>> out at brower side looking at firebug. >> >>>>>>> >> >>>>>>> On 9/27/11, Massimo Di Pierro <massimo.dipie...@gmail.com> wrote: >> >>>>>>>> Other than the error, do you see wrong behavior from the browser? >> >>>>>>>> I do not not believe that there is anything wrong as we ship the >> >>>>>>>> "official" rocket. We are just printing an exception that before >> >>>>>>>> was >> >>>>>>>> being ignored. >> >>>>>>>> >> >>>>>>>> Massimo >> >>>>>>>> >> >>>>>>>> On Sep 26, 2:47 pm, Phyo Arkar <phyo.arkarl...@gmail.com> wrote: >> >>>>>>>>> Hello Web2py. >> >>>>>>>>> >> >>>>>>>>> Rocket server was very well stable for me until today , after >> >>>>>>>>> 2 >> >>>>>>>>> weeks of update and while testing usual Json output which was >> >>>>>>>>> never >> >>>>>>>>> changed and working fine fore 8 months. >> >>>>>>>>> >> >>>>>>>>> This happens! >> >>>>>>>>> >> >>>>>>>>> ERROR:Rocket.Errors.Thread-6:Unhandled Error when serving >> >>>>>>>>> connection: >> >>>>>>>>> Traceback (most recent call last): >> >>>>>>>>> >> >>>>>>>>> File "/root/web2py/gluon/rocket.py", line 1064, in run >> >>>>>>>>> self.run_app(conn) >> >>>>>>>>> >> >>>>>>>>> File "/root/web2py/gluon/rocket.py", line 1552, in run_app >> >>>>>>>>> self.write(data, sections) >> >>>>>>>>> >> >>>>>>>>> File "/root/web2py/gluon/rocket.py", line 1472, in write >> >>>>>>>>> self.send_headers(data, sections) >> >>>>>>>>> >> >>>>>>>>> File "/root/web2py/gluon/rocket.py", line 1456, in >> >>>>>>>>> send_headers >> >>>>>>>>> self.conn.sendall(b(header_data)) >> >>>>>>>>> >> >>>>>>>>> File "/usr/lib/python2.7/socket.py", line 224, in meth >> >>>>>>>>> return getattr(self._sock,name)(*args) >> >>>>>>>>> >> >>>>>>>>> error: [Errno 32] Broken pipe >> >>>>>>>>> >> >>>>>>>>> and it is on and off randomly. >> >>>>>>>>> >> >>>>>>>>> Why? >> >>>>>>>>> >> >>>>>>>>> heres my Json Controller : >> >>>>>>>>> >> >>>>>>>>> def rows(): >> >>>>>>>>> >> >>>>>>>>> if REQUEST.vars.db_name: >> >>>>>>>>> casesdb = DAL( 'mysql://root@localhost/' + >> >>>>>>>>> REQUEST.vars.db_name ) >> >>>>>>>>> casesdb.define_table( 'email_data', migrate = >> >>>>>>>>> False, >> >>>>>>>>> *email_halfschema ) >> >>>>>>>>> casesdb.define_table( 'loosefile_data', migrate >> >>>>>>>>> = >> >>>>>>>>> False, >> >>>>>>>>> *file_halfschema ) >> >>>>>>>>> casesdb.define_table( 'attach_data', migrate = >> >>>>>>>>> False, >> >>>>>>>>> *file_halfschema ) >> >>>>>>>>> elif SESSION.db_name: >> >>>>>>>>> casesdb = DAL( 'mysql://root@localhost/' + >> >>>>>>>>> SESSION.db_name >> >>>>>>>>> ) >> >>>>>>>>> casesdb.define_table( 'email_data', migrate = >> >>>>>>>>> False, >> >>>>>>>>> *email_halfschema ) >> >>>>>>>>> casesdb.define_table( 'loosefile_data', migrate >> >>>>>>>>> = >> >>>>>>>>> False, >> >>>>>>>>> *file_halfschema ) >> >>>>>>>>> casesdb.define_table( 'attach_data', migrate = >> >>>>>>>>> False, >> >>>>>>>>> *file_halfschema ) >> >>>>>>>>> >> >>>>>>>>> fields = ['id', 'filePath', 'fileName', 'cus', >> >>>>>>>>> 'sentto',\ >> >>>>>>>>> 'emailFrom', 'subject', 'cc', 'bcc', >> >>>>>>>>> 'extracted', >> >>>>>>>>> 'hdrs',\ >> >>>>>>>>> >> >>>>>>>>> >> >>>>>>>>> >> >>>>>>>>> 'DateTime','OriginalDateTime','TimeZone','reply_to',"master",'duplicated',' >> >>>>>>>>> MD5Hash','msgID','fqdn'] >> >>>>>>>>> rows = [] >> >>>>>>>>> page = int( request.vars.page ) >> >>>>>>>>> >> >>>>>>>>> max = request.vars.totalrows >> >>>>>>>>> if max: >> >>>>>>>>> pagesize = int( max ) >> >>>>>>>>> else: >> >>>>>>>>> pagesize = int( request.vars.rows ) >> >>>>>>>>> print pagesize >> >>>>>>>>> limitby = ( page * pagesize - pagesize, page * pagesize >> >>>>>>>>> ) >> >>>>>>>>> # limitby = ( 1, 25 ) >> >>>>>>>>> print "str( page * pagesize - pagesize )" + str( page * >> >>>>>>>>> pagesize >> >>>>>>>>> - >> >>>>>>>>> pagesize ) + " limitby " + str( page * pagesize ) >> >>>>>>>>> # orderby = request.vars.sidx >> >>>>>>>>> orderby = casesdb.email_data[request.vars.sidx] >> >>>>>>>>> print orderby >> >>>>>>>>> if request.vars.sord == 'desc': >> >>>>>>>>> orderby = ~orderby >> >>>>>>>>> >> >>>>>>>>> query = casesdb.email_data.id > 0 >> >>>>>>>>> >> >>>>>>>>> for r in casesdb( query ).select( orderby = orderby, >> >>>>>>>>> limitby >> >>>>>>>>> = >> >>>>>>>>> limitby ): >> >>>>>>>>> vals = [] >> >>>>>>>>> for f in fields: >> >>>>>>>>> if f == 'extracted': >> >>>>>>>>> vals.append( _sentisize(r[f]) ) >> >>>>>>>>> else : >> >>>>>>>>> vals.append( r[f] ) >> >>>>>>>>> rows.append( dict( id = r.id, cell = vals ) ) >> >>>>>>>>> >> >>>>>>>>> total = casesdb( query ).count() >> >>>>>>>>> print total >> >>>>>>>>> if total % pagesize >0: >> >>>>>>>>> pages = int( total / pagesize ) +1 >> >>>>>>>>> else: >> >>>>>>>>> pages = int( total / pagesize ) >> >>>>>>>>> #if total % pagesize == 0: pages -= 1 >> >>>>>>>>> data = dict( total = pages, page = page, rows = rows ) >> >>>>>>>>> return data >> >>>>>>> >> >>>>>> >> >>>>> >> >>>> >> >>> >> >> > >
