a lot of crit error in logs

[Maxim Dounin]

Hello!

On Fri, Apr 16, 2010 at 10:28:10AM -0400, solarisik wrote:

> I use nginx for reverse proxy on front of apache2. I get a lot of crit messages in logs like this.
>  
> 2010/04/16 17:20:32  1730#0: *468039 rename() "/webhost/tmp/0004372283" to "/webhost/tmp/xxxxxxx/" failed (20: Not a directory) while reading upstream, client: xxxxxxxxxxx, server: *.*, request: "GET /xxxxxxxx/?&manset_id=&baslangic=2510 HTTP/1.1", upstream: "http://xxxxxxxxxxx:8080/xxxxxxxxxxxx/?&manset_id=&baslangic=2510", host: "xxxxxxxxxx.com"
> 
> here is my config.

[...]

>                 root /webhost/tmp/;

[...]

> proxy_store on;
> proxy_redirect off; 
> proxy_temp_path /webhost/tmp; 

You are configured proxy_store to /webhost/tmp/.  And you have 
other files in this directory, at least nginx's own proxy_temp 
and/or other files already saved by previous proxy_store requests 
(which is likely to cause conflicts due to your site structure).

Additionally, you are trying to save directory index accesses 
without mapping them to files via custom proxy_store strings.  
This will not work (though will cause different error).

And, after all, you don't use proxy_store results at all.

Let me guess: you don't need proxy_store at all, right?

Maxim Dounin