Il giorno 23/set/2010, alle ore 14.39, Jean-Baptiste Quenot ha scritto: > This is the nginx access log format I use: > > log_format mine '$remote_addr $remote_user $bytes_sent > $request_time $upstream_response_time ' > '"$request" $status $body_bytes_sent "$http_referer" ' > '"$http_user_agent" "$http_x_forwarded_for" "$server_name"'; > > You'll find the errors of the day at > http://dl.dropbox.com/u/5836407/uwsgi_errors.txt > -- >
Bingo :) As i have said, uWSGI is responding to connection request (so the time between connect() and accept()) very slowly, your response time are all near 30 (29.xxx) but no so slow to timeout nginx. How to solve this ? This is hard to answer, adding another couple of workers could solve the problem, but the rate of error request is so little that consuming more memory or cpu resource could not be a good idea. Check if there are slow part of your code that can be managed by grunt processes instead of workers. Pratically when a request is potentially slow you can detach it from the pool and spawn another worker while the old request continue its job. (good candidates are image processing code, big queries and connection to external servers) -- Roberto De Ioris http://unbit.it JID: [email protected] _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
