Re: Problems with cache misses when using Squid in httpd accelerator mode

Danny Cecil wrote:
>
> Does anyone have any links to technical information or white papers on
> using Squid as an httpd accelerator?

It is technically the same thing as running it as a proxy, only that you
have control of all parts, and need some additional configuration to
tell Squid how to find the destination server to proxy the request to.

> Since our site is relatively small, and we have a
> large cache configured for Squid (3 Gb), it seems to me that once an
> item is loaded into cache, it should be fetched from cache until it
> ages out.

How long a entry is considered fresh depends on a number of things,
mainly what kind of expiry information your server is providing to Squid
for the requests, and how your users are behaving with that reload
button..

> I?ve been pulling my hair out trying to debug this problem
> and have not found a lot of information on configuring/tuning Squid as
> a reverse proxy. Any pointers or ideas that you have would greatly
> appreciated!

The only tunings relevant in Squid are:

* refresh_pattern for adding expiry information (better done at the
origin server).

* The various options for ignoring/changing users Reload requests. Squid
can do three things with a reload request:
  a) Act on it and bypass the cache
  b) Transform it to a If-Modified-Since request to the origin server if
the object is cached
  c) Ignore it and trust your expiry setup.

For an accelerator you probbaly want either b or c depending on how well
you have configured your web server(s) to signal expiry information to
Squid.

Search for reload in squid.conf.default to find the available options.
There are both global options and options to refresh_pattern for more
fine grained control.

--
Henrik Nordstrom
Squid hacker