On 30/07/2013 1:49 a.m., ana any wrote:

hi guys,

I have squid-3.1.20 installed on ubuntu-12.10. I didn't change much in
squid's default configurations. I just defined "cache_dir" and
"nameservers" and enabled "localnet". now when I open any webpages from
client, in my access.log I see lots and lots of this message (for almost all 
contents):

...

1375088677.969   3341 192.168.0.81 TCP_MISS/200 111920 GET http://www.imdb.com/ 
- DIRECT/72.21.215.52 text/html
...
1375088683.295    479 192.168.0.81 TCP_MISS/200 648 GET 
http://s.media-imdb.com/twilight/? - DIRECT/72.21.215.52 image/gif
....


I look at squid's cache folder and I see size of cache folder grows (so content 
is being
cached?) but apparently no cached content is being used.

In direct accordance with IMDB cache control rules about what may be done with their website.

The first of those URLs presents an index page containing proprietary and copyright infromation under a limited license. It also contains embeded user-specific content. It is marked as *private* information for the individual client and may not be stored by shared caches such as Squid.

The second URL is unclear exactly since you have cropped off all the query-string portion which would aid us in identifying the exact problem. Other similar URLs are marked with "no-cache, must-revalidate" and an Expiry timestamp long in the past. To cause HTTP/1.0 software (such as squid-3.1) not to cache or to revalidate on every use, and HTTP/1.1 compliant softwrae (such as squid-3.2 and later) to cache it but revalidate before any use.

I googled but didn't find anything that help me. is there anything I'm missing? 
Does anyone have a thought or suggestion?

Best thing to do is upgrade to 3.2 or 3.3.

There are refresh_pattern directive options which can fine tune behavioru for unfriendly site. However I don't advise doing anything with those particular URLs due to the embeded user-specific content IMDB adds to their pages (you do not want to be responsible for your users getting each others IMDB account details or notices). If the problem is more widespread in other sites though you may want to share your whole config file (without the # comment lines) and we may be able to find something to help.

Amos

Reply via email to