Re: [squid-users] inconsistent caches using a sibling cache hierarchy

From: Domingos Parra Novo <domingosnovo@dont-contact.us>
Date: Tue, 30 May 2006 09:33:40 -0300

        Hiyas,

Matus UHLAR - fantomas wrote:
> On 29.05.06 22:20, Domingos Parra Novo wrote:
>> I'm using a pool of (4) squid servers as a web accelerator for a slow
>> backend (vignette, to be more exact). Right now, almost everything works
>> like a charm, besides one thing.
>
>> If, for any reason (high load, for example), a purge request (for object
>> "foo") reaches servers 1, 2 and 3 (but haven't got the time to purge the
>> object on server 4), and a new request for the object "foo" arrives on
>> server 1 (which already expired this object), the request is redirected
>> from server 1 to server 4 (which contains an old version of the object).
>> In a few words, I sometimes get a invalid cache, with old objects on my
>> squid servers.
>
> does your backend web server provide correct expiry informations? Does it
> provide correct Cache-Control: headers?

        Yes, it does, tough I forgot to mention that I disabled IMS reloads on
my squid farm. My vignette farm is already overloaded (and thats why I'm
trying to cache content on squid). If I let the user "check" if the
content is fresh or not, by simply hitting Ctrl+F5, my backend would die
a painless death in 5 minutes or so.

> I think the main problem is that your caches provide old content, it has
> nothing to do with their ICP communication.

        Well, in truth, my squid farm does exactly what they where meant to do.
Ie, always try to get a content from its sibling caches, and only if
this fails, retrieve it on the backend server.

        I was just trying to find a way to make the cache a little "smarter".
Forcing a recently "purged/stale" object to stay in cache for 30 more
seconds (and not giving ICP_HITs on this period) would be more then enough.

>> By the way, would htcp help me on this task? I know its "smarter" then
>> ICP, but I haven't seen much documentation about any of them, to tell you
>> the truth.
>
> I don't think so, however HTCP is not fully supported in squid and afaik
> it currently provides no more functionality than ICP.

        So, there's no advantage on using HTCP on squid 2.5 right now? Having
the same subset of functionality I'd get on ICP, and with a larger UDP
packet would be bad, IMHO. :)

Thanks in advance,

Domingos.

-- 
Domingos Parra Novo
Coordenador de Projetos
Terra Networks Brasil S/A
Tel: +55(51)3284-4275
Received on Tue May 30 2006 - 06:33:43 MDT

This archive was generated by hypermail pre-2.1.9 : Thu Jun 01 2006 - 12:00:02 MDT