Going live!

[Igor Sysoev]

On Mon, Feb 09, 2009 at 03:24:27PM +0100, Andreas Andersson wrote:

> We just turned on nginx and started directing our customers through it and
> it seems to be working fine. However, I have some questions.
> 
> I started monitor the error.log as soon as we started and I get loads of
> stuff there. Generally I expect only errors to go to the error log but this
> doesn't seem to be true for nginx. These are the two messages I get all the
> time
> 
> 1) xxx.xxx.xxx.xxx closed keepalive connection (104: connection reset by
> peer).
> As far as I understand from previous mails to this list this is just
> something you'll get on EVERY request from IE. True?
> 
> 2) an upstream response is buffered...
> This too seems to be fine, just that the file is bigger than what can be
> buffered in memory and instead is written to disk before being sent.
> 
> This seems to be good stuff so why are the written to the error log?

The "keepalive" messages are logged on info level, while "buffered" are
logged on warn level. Set

     error_log   /path/to/log  error;

to see error, crit, alert, and emerg messages.


-- 
Igor Sysoev
http://sysoev.ru/en/