>>> Hello. >>> >>> I've got two instance in amazon cloud. On the first instance i've got >>> my python app and on the second instance I run siege (or apache >>> benchmark). If my app return response with big size (~200kb) response >>> time = 10 sec, if size of response less then 1 kb or equal response >>> time is 0.20 sec. I run siege in amazon intranet. >>> Maybe, I should run uwsgi with some options (maybe buffer) or somehow >>> tune nginx? >>> >>> Please help. > >> please report your uWSGI configuration > > [uwsgi] > socket = 0.0.0.0:9122 > processes = 16 > harakiri = 1 > harakiri-verbose = true > daemonize = /var/log/app/uwsgi.log > master = true > #buffer-size = 65536 > pidfile = /var/run/app.pid > no-orphans = true > no-default-app = true > vacuum = true > module = myapp.server > callable = app > memory-report = true > touch-reload=/usr/share/ > pyshared/app > post-buffering = 2048 # - this is not necessary > post-buffering-bufsize = 262144 # - this is not necessar >
Hi, your config is good, maybe you are reaching some system limit. It depends on how many concurrent requests you are sending. With your setup you can manage 16 concurrent requests, so with a 200k stream you will generate about 3.2 megs. If you smash your server you can easily increase response times. I suggest you (when you do stress-test), to start always with a single uWSGI process and then increase until you find the best combination for your hardware/app -- Roberto De Ioris http://unbit.it _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
