400 response ...

[Alan Williamson]

interesting, i had to restart the service here to get this logging 
enabled; a simply HUP didn't work.

okay, back to this though -- what makes nginx throw a 400 response?


in the error log i get lots of these:

2008/07/14 08:22:09 [info] 9392#0: *54478 client 91.149.53.139 closed 
keepalive connection
2008/07/14 08:22:09 [info] 9391#0: *55972 client 193.252.122.88 closed 
keepalive connection
2008/07/14 08:22:09 [info] 9391#0: *55981 client 88.191.15.62 closed 
keepalive connection


some of these:

2008/07/14 08:22:10 [warn] 9392#0: *56100 an upstream response is 
buffered to a temporary file /usr/local/nginx/proxy_temp/1/35/0000000351 
while reading upstream, client: 87.122.18.68, server: 
rss.feedsportal.com, request: "GET /c/269/f/3544/index.rss HTTP/1.1", u


very few of these:
2008/07/14 08:23:53 [info] 9392#0: *105147 client timed out (110: 
Connection timed out) while reading client request line, client: 
213.186.114.166, server: 0.0.0.0:80


Yet we get TONS of 400's being reported in the access logs.

What happens if a backend isn't able to process a connection?  Does 
nginx not through a 503 instead?


Igor Sysoev wrote:
> On Fri, Jul 04, 2008 at 03:50:47PM +0100, Alan Williamson wrote:
> 
>>>> We are using nginx as a frontend loadbalancer, but our code does not 
>>>> generate 400 errors.  So under what circumstances would nginx return a 
>>>> 400 request?
>>> Reasons of these errors are logged at info level in error_log:
>>>
> The default level is notice.
> Unimportant client errors are logged at info level.