[squid-users] caching and changing content

Amos Jeffries squid3 at treenet.co.nz
Thu Sep 12 09:37:41 UTC 2019

On 12/09/19 8:25 pm, fansari wrote:
> In my scenrio (squid 3.5.23) I have several clients which download content.
> Now I want to achieve two things: if one client has already downloaded the
> content the second client requesting the same content should take it from
> the squid cache.
> But: when this content resource on the internet is changing the squid should
> not give a HIT but a MISS and refresh the content the next time a client is
> requesting it.

No. HTTP specifies that the cache should update its copy of the content,
and deliver the new result. This is *not* a MISS. It is a REFRESH in the
logs, or a "Near-HIT" in the traffic statistics report.

> It is not clear to me how to achieve this behaviour.

You do not have to do anything. For content which this behaviour is
possible, Squid does it by default.

> The application (server on internet) sends headers like this:
> cache-control: public, max-age=0

'public' - this object is allowed to be stored in caches.

'max-age=0' - the cache is required to contact the server for updates
before delivering this object to any clients.

> last-modified: Thu, 12 Sep 2019 07:30:56 GMT
> etag: W/"6699e3-16d2461dba1"
> So far I could only achieve HITs using refresh_pattern with min value > 0.
> refresh_pattern .               30       20%     1440 override-expire

refresh_pattern provides default values for Squid to use in its caching
algorithm if the server does not provide them.

The headers you show above contain clear and precise cache-control
instructions. So refresh_pattern is not normally relevant to that object.

By adding the override-expires option you are *forcing* the object to be
cached for no less than 30 minutes and no more than 1 week...

> But also this configuration is not exactly what  I want. 
> Now even if the content of a resource on the internet is changing I get HITs
> for 30 min.

Not just that resource. By using the '.' pattern you are forcing this
30min HIT on *every* object any client fetches through this cache.

It is best to be a specific as you can in the pattern field. Especially
when using the ignore-* and override-* options. To reduce nasty side

PS. The server is delivering cache controls, what makes you think the
server developer and admin are wrong about their own content and systems?


More information about the squid-users mailing list