Error with nginx cache: no space left on /tmp

[clopezi]

Hi,

Today our server did something strange, but i solved. I make the post for one question and maybe will be helpfull to other user.

Suddenly, the server make a error 500, after that, i restart all services but varnish cache don't restart. When i go to check the varnish configuration, the script says: no space left in /tmp/

I check with df -h and /tmp have around 2GB free (32% of space).

I remove all session* in /tmp and all in /tmp/nginx_client and now all works perfectly, with a 90% empty in /tmp

So i have two questions:

1) Why no space empty if df -h says yes?

2) Who can i make a cronjob to empty nginx dir and after that restart nginx service every night? The following command in crontab is correct?

0 0 * * * rm -rf /temp/nginx_client/* && /etc/init.d/nginx restart

Thank you!

 

[Jeff Shotnik]

Nginx has a cache manager process that watches the nginx cache and removes inactive files after a certain period or when the cache limit has been reached. In order to keep /tmp from becoming full you'll need to adjust your 'proxy_cache_path' directive to use a smaller cache size. If /tmp is too small for the cache size you need, move nginx cache to /home, or a larger partition.

Here's a link on how to use proxy_cache_path - HttpProxyModule

"Why no space empty if df -h says yes?"
>> You may need to run `sync` to get the correct value to show - changes aren't committed until buffers are flushed.