> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA512
>
> Using the following simple code:
>
> import uwsgi
>
> def KV55(environ, start_response):
> response = environ['wsgi.input'].read(int(environ['CONTENT_LENGTH']))
> start_response('200 OK', [('Content-Type', 'text/xml'),
> ('Content-length', str(len(response)))])
> yield response
>
> uwsgi.applications = {'':KV55}
>
>
> I am getting a one second penalty timeout because I am reading the
> wsgi.input.
>
> To test this I am using:
>
> uwsgi --http 127.0.0.1:8181 -w simple
> curl -X POST -T kv55.xml http://127.0.0.1:8181
>
> uwsgi gives me:
>
> [pid: 6625|app: 0|req: 1/1] 127.0.0.1 () {30 vars in 400 bytes} [Tue
> Jan 24 02:46:41 2012] POST /kv55.xml => generated 3935 bytes in 1001
> msecs (HTTP/1.1 200) 2 headers in 65 bytes (1 switches on core 0)
>
>
>
> When simple not touching the wsgi.input at all:
>
> [pid: 6702|app: 0|req: 1/1] 127.0.0.1 () {30 vars in 400 bytes} [Tue
> Jan 24 02:49:53 2012] POST /kv55.xml => generated 0 bytes in 0 msecs
> (HTTP/1.1 200) 2 headers in 62 bytes (1 switches on core 0)
>
>
> When I am specifing --post-buffering 1 the situation gets worse even
> when the wsgi.input is not touched at all.
>
>
> Obviously I know how to debug; strace shows me:
>
> writev(8, [{"\0\220\1\0", 4}, {"\16\0", 2}, {"REQUEST_METHOD", 14},
> {"\4\0", 2}, {"POST", 4}, {"\v\0", 2}, {"REQUEST_URI", 11}, {"\t\0",
> 2}, {"/kv55.xml", 9}, {"\t\0", 2}, {"PATH_INFO", 9}, {"\t\0", 2},
> {"/kv55.xml", 9}, {"\f\0", 2}, {"QUERY_STRING", 12}, {"\0\0", 2}, {"",
> 0}, {"\17\0", 2}, {"SERVER_PROTOCOL", 15}, {"\10\0", 2}, {"HTTP/1.1",
> 8}, {"\v\0", 2}, {"SCRIPT_NAME", 11}, {"\0\0", 2}, {"", 0}, {"\v\0",
> 2}, {"SERVER_NAME", 11}, {"\7\0", 2}, {"nemesis", 7}, {"\v\0", 2},
> {"SERVER_PORT", 11}, {"\4\0", 2}, ...], 65) = 404
> epoll_ctl(6, EPOLL_CTL_MOD, 8, {EPOLLIN, {u32=8, u64=8}}) = 0
> epoll_wait(6, {{EPOLLIN, {u32=7, u64=7}}}, 64, 30000) = 1
> recvfrom(7, "<?xml version=\"1.0\"?>\n\n<DRIS_55>"..., 65536, 0, NULL,
> NULL) = 3935
>
> The epoll_wait is actually where the time is spent.
>
>
> Could anyone help me out, getting a better response time?
>
>
Expect 100-continue has been implemented in latest tip. As it is a feature
you may want to implement in the app itself (instead of the http
server/router), you have to enable it with --http-manage-expect
--
Roberto De Ioris
http://unbit.it
_______________________________________________
uWSGI mailing list
[email protected]
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
