I'm using nginx + php-fpm, but there is abnormal load on /tmp when no files are noted in /tmp at all.

[Runner Up]

I'm using CentOS release 5.5 (Final)
nginx-0.9.3
remi yum rpms for php-fpm, php, and php-apc

My php-fpm setting are near default except for spawn child numbers.

The server is used for no other purpose other than niginx + php-fpm and gets
about 2 millions hits on a very heavy duty php script per day.
(DB server is seperate)

iostat shows
Device:         rrqm/s   wrqm/s   r/s   w/s   rsec/s   wsec/s avgrq-sz
avgqu-sz   await  svctm  %util
sda               0.00   107.40  0.00 175.80     0.00  2268.80    12.91
110.18  632.36   5.69 100.04
....
sda5              0.00   104.60  0.00 174.40     0.00  2233.60    12.81
109.25  631.62   5.74 100.04
...
sdb               0.00    85.40  0.00  2.20     0.00   700.80   318.55
0.01    3.18   0.82   0.18
sdb1              0.00    85.40  0.00  2.20     0.00   700.80   318.55
0.01    3.18   0.82   0.18

sdb is the hard drive with the web files and sda5 is /tmp

ls -alh /tmp
total 40K
drwxrwxrwt  5 root root 4.0K Feb 18 00:21 .
drwxr-xr-x 24 root root 4.0K Feb 15 02:27 ..
drwxrwxrwt  2 root root 4.0K Jan 18 01:58 .font-unix
drwxrwxrwt  2 root root 4.0K Jan 18 01:58 .ICE-unix
drwx------  2 root root  16K Aug 26  2009 lost+found

If nginx is stopped the load on sda5 goes away.

Is there any reason /tmp should have such a heavy write load without any
files in it?

nginx virtual host conf
location ~ \.(php|html|htm)$ {
     proxy_read_timeout 180;
     proxy_connect_timeout 120;
     proxy_send_timeout 180;
     proxy_buffers 32 4k;
     proxy_max_temp_file_size 0;
     fastcgi_pass 127.0.0.1:9000;
     fastcgi_index index.php;
     fastcgi_param SCRIPT_FILENAME /[DELETED MY HOME]$fastcgi_script_name;
     include /usr/local/nginx/conf/fastcgi_params;
  }
fastcgi_params
default values with the following
fastcgi_connect_timeout 60;
fastcgi_send_timeout 180;
fastcgi_read_timeout 180;
fastcgi_buffer_size 128k;
fastcgi_buffers 4 256k;
fastcgi_busy_buffers_size 256k;
fastcgi_temp_file_write_size 256k;
fastcgi_intercept_errors on;
fastcgi_max_temp_file_size 0;

Thanks.

My question was :
Is there any reason /tmp should have such a heavy write load without any
files in it?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://nginx.org/pipermail/nginx/attachments/20110218/5141975f/attachment.html>