Re: [squid-users] Sibling A does not respond with sibling_hit even though there is a refreshed item in the sibling B (Squid 3.1)

From: Michael Pye <michael_at_ulimit.org>
Date: Sun, 11 May 2014 13:37:02 +0100

On 25/03/2013 05:55, sugusbs wrote:
> 1. Sibling A receives 1st request for R and responds with tcp_miss and
> FIRST_UP_PARENT.
> 2. Sibling B receives 2nd request for R and responds with sibling_hit
> due to sibling A.
> 3. After R object gets expired, sibling A receives 3rd request for R and
> responds with tcp_miss and FIRST_UP_PARENT.
> 4. Sibling B receives 4th request for R and should respond with
> sibling_hit but tcp_miss and FIRST_UP_PARENT.
>
> The object R has been refreshed in step 3, why does sibling B responds with
> FIRST_UP_PARENT in step 4?

I have the same question, using squid 3.1 It seems that whenever an
objects exists in the cache but is stale, other siblings are not checked
for a fresher object but only the origin server.

Is it possible to force a check of the object via a sibling first ?

if the object doesn't exist siblings are queried correctly, it's only
when an object exists but is stale.

Michael
Received on Sun May 11 2014 - 12:37:27 MDT

This archive was generated by hypermail 2.2.0 : Sun May 11 2014 - 12:00:04 MDT