Connection reset errors

Server NGINX
1

Hello,

I’m using nginx to proxy for both ruby on rails and php applications.
Some of my users are reporting intermittant connection reset errors in
their browsers. The behavior they describe consists of the browser fully
loading the web page then immediately forwarding them to a connection
reset page. I have seen it first hand a couple of times but have not
been able to consistently reproduce it. I checked the access log for the
request and it never gets logged by nginx. Has anyone else seen this
kind of behavior?

My guess is its some kind of bug with 302 handling.

Thanks.

2

Adam M. ha scritto:

My guess is its some kind of bug with 302 handling.

You should report the content of the error log.

Thanks.

Regards Manlio P.

3

We have seen this behavior when connecting directly to mongrel. We
suspect it is a mongrel bug, but we’ve done no investigation.

kr

4

The request is showing in the access log but not in the error log when
set
to info.

----- Original Message -----
From: “Igor S.” [email protected]
To: [email protected]
Sent: Saturday, October 13, 2007 2:58 AM
Subject: Re: connection reset errors

5

On Sat, Oct 13, 2007 at 05:03:22AM -0400, Adam M. wrote:

The request is showing in the access log but not in the error log when set
to info.

So there were no errors on system level (Connection reset by peer).
The debug log may help to resolve the problem:

./configure --with-debug …

nginx.conf:

error_log /path/to/log debug;

6

On Fri, Oct 12, 2007 at 02:07:07PM -0400, Adam M. wrote:

I’m using nginx to proxy for both ruby on rails and php applications. Some of my users are reporting intermittant connection reset errors in their browsers. The behavior they describe consists of the browser fully loading the web page then immediately forwarding them to a connection reset page. I have seen it first hand a couple of times but have not been able to consistently reproduce it. I checked the access log for the request and it never gets logged by nginx. Has anyone else seen this kind of behavior?

My guess is its some kind of bug with 302 handling.

Could you set nginx error_log to info level:

error_log /path/to/log info;

reproduce the bug and show the error_log ?