On 11/06/2013 11:32 p.m., Alex Domoradov wrote:
On Tue, Jun 11, 2013 at 11:21 AM, Eliezer Croitoru wrote:
On 6/11/2013 10:56 AM, Alex Domoradov wrote:
Amos, any idea?
Just asking.
do you understand the difference between a TCP_HIT and
TCP_REFRESH_UNMODIFIED ??
Looking at the wiki again it seems like it would be helpful just to see this
specific section:
http://wiki.squid-cache.org/SquidFaq/SquidLogs?action=fullsearch&context=180&value=TCP_REFRESH_UNMODIFIED&titlesearch=Titles#Squid_result_codes
You might not understand that:
TCP_REFRESH_UNMODIFIED is actually a HIT based on the origin server(or a
parent proxy).
if you have lots of TCP_REFRESH_UNMODIFIED in your log file you can tell
your clients (in a case of only one proxy between the client and the site)
that they get it from cache while making sure it's the file they should get
and also the server knows that yet another client gets a good and valid
reply.
try to make a calculation of all these requests like this:
sum of bytes that was sent to clients less the bytes that was sent on the
304 request from squid = you got your HIT in bandwidth length rather then
FULL not From Wire transfer.
But in such case (with reload-into-ims and refresh_pattern) should
such requests exists at all? All I want is to ignore ALL caching
options and timestamp and always get objects directly only from cache
Yes they should exist in that case. "reload-into-ims" just means convert
client "reload" requests into IMS requests for the origin.
Amos
