On 19/10/11 07:02, goran kent wrote:
Hi,
Using squid 2.6.STABLE21 on CentOS 5.x as a reverse proxy.
For our application caching URLs with "/?q=" in it is vital for
performance. In other words, irrespective of client IP or client
cache settings (or even origin server cache control), etc, any URL
with "/?q=" in it *must* be cached permanently and served to any and
all clients (until we refresh it manually).
The problem I'm having is that clientA will browse to eg,
"test.com/?q=a" and it will be cached successfully. However, clientB
or clientC, etc, browsing to the same URL ("test.com/?q=a" in this
case), will not always be served from the cache. Instead, and I've
been unsuccessfully trying to isolate why, squid will go direct and
re-fetch it from the origin server, even though the page in the cache
is only a few seconds old.
Squid is optimized to produce an accurate copy of the object as per the
client request requirements.
Here's the log:
ip.of.my.desktop TCP_MISS/200 3211 GET http://test.com/?q=minors -
FIRST_UP_PARENT/myAccel text/html
ip.of.client.A TCP_MISS/200 3258 GET http://test.com/?q=minors -
FIRST_UP_PARENT/myAccel text/html
ip.of.client.B TCP_MISS/200 3206 GET http://test.com/?q=minors -
FIRST_UP_PARENT/myAccel text/html
Browsing again to the same URL with clientA yields a hit:
ip.of.client.A TCP_MEM_HIT/200 3214 GET http://test.com/?q=minors -
NONE/- text/html
So, it's almost like caching is tied to the client IP, and not based
solely on the cached URL, which is what I would expect for reverse
proxying.
It is tied to the client request headers usually.
Grab a copy of the request and reply headers and find out.
I'd appreciate any comments/suggestions on what I'm missing.
Thanks
Henry
My config:
http_port 80 accel defaultsite=test.com
cache_peer directtest.com parent 80 0 no-query originserver name=myAccel
acl our_sites dstdomain test.com
acl all src 0.0.0.0/0.0.0.0
http_access allow our_sites
cache_peer_access myAccel allow our_sites
cache_peer_access myAccel deny all
# my next question will be: what's the best refresh_pattern line to
permanently cache pages containing my pattern:
refresh_pattern -i [?]q= 1000000 100 100 reload-into-ims
override-expire override-lastmod ignore-reload ignore-no-cache
That requests line says store for a maximum of 100 seconds regardless of
when the origin server says it will change.
Start with an upgrade. The newer Squids have much better reverse proxy
support. Then work from there.
Amos
--
Please be using
Current Stable Squid 2.7.STABLE9 or 3.1.16
Beta testers wanted for 3.2.0.13
