As mentioned caching in the frontend is disabled, but backends are still
recording only misses (TCP_MISS/200, TCP MISS/204, TCP_MISS/304) and
almost all requests are passed to the origin server
(HIER_DIRECT/origin-server-ip).
After recompiling squid with --enable-cache-digests frontend is
recording CD_PARENT_HIT no more, instead only CARP/backend-1 or backend-2.
These are in general frontend and backend configuration files:
frontend.conf
------------------
include /some/common/acl
http_port 3128
cache_peer localhost parent 4001 0 carp name=backend-1login=PASSproxy-only
cache_peer localhost parent 4002 0 carp name=backend-2 login=PASSproxy-only
cache_mem 768 MB
refresh_pattern -i \.(jpg|jpeg|gif|png|swf|htm|html|bmp)(\?.*)?$ 21600
100% 21600 reload-into-ims ignore-reload ignore-no-cache ignore-auth
ignore-private
refresh_pattern ^ftp: 1440 20% 10080
refresh_pattern ^gopher: 1440 0% 1440
refresh_pattern -i (/cgi-bin/|\?) 0 0% 0
refresh_pattern . 0 20% 4320
shutdown_lifetime 3 second
backend.conf
------------------
include /some/common/acl
http_port 127.0.0.1:4001
visible_hostname backend-1
unique_hostname backend-1
cache_dir aufs /Cache1 1024 128 128
cache_mem 8 MB
refresh_pattern -i \.(jpg|jpeg|gif|png|swf|htm|html|bmp)(\?.*)?$ 21600
100% 21600 reload-into-ims ignore-reload ignore-no-cache ignore-auth
ignore-private
refresh_pattern ^ftp: 1440 20% 10080
refresh_pattern ^gopher: 1440 0% 1440
refresh_pattern -i (/cgi-bin/|\?) 0 0% 0
refresh_pattern . 0 20% 4320
cache_swap_low 90
cache_swap_high 95
maximum_object_size 512 MB
Artemis
On 01/07/2011 04:39 PM, Amos Jeffries wrote:
On 07/01/11 23:29, Artemis BRAJA wrote:
As suggested, I upgraded to 3.2.0.4 with --disable-cpu-profiling option
but I'm still getting TCP_MISS/200 on both backends and I can't see any
TCP_HIT.
Artemis
CD_PARENT_HIT is a HIT instead of TCP_HIT. Though it indicates a
cache-digest hit and I don't see digest enabled in your example config.
Note: If you have disabled caching in the frontend like most CARP
configs need you will not see any HIT there as all requests are MISS
at frontend and passed to backend.
The backend logs should still be recording HIT/MISS like normal.
On 01/07/2011 05:36 AM, Amos Jeffries wrote:
On 06/01/11 04:54, Artemis BRAJA wrote:
Hello everyone!
I recently upgraded to squid 3.2.0.2 and started successfully
configuring multiple squid instances on a dual core linux box with two
frontend squid handling authentication and load balancing and also two
backend instances handling caching.
Frontend-1---------Backend-1
/ \ /
/ \/
Client ------> / \
\ / \
\ / \
Frontend-2---------Backend-2
Both frontend instances are configured as following:
cache_peer localhost parent 4001 0 carp login=PASS name=backend-1
cache_peer localhost parent 4002 0 carp login=PASS name=backend-2
Load balancing between frontends (iptables) and backends (CARP) is
working fine, but I see only TCP_MISS/200, TCP_MISS\302 and
CD_PARENT_HIT/backend-1-2 on frontends and backends access.log.
Is this normal when squid is configured this way or there is some
misconfiguration on squids config file?
Please use 3.2.0.4 now (requires --disable-cpu-profiling option to fix
a regression).
You may have hit the INM bug which increased MISS rates or be seeing a
lot of mixed Chrome and non-Chrome traffic. Chrome has recently
started using a new encoding format which makes its Accept-Encoding
headers different to other browsers. The result is more TCP_MISS/200
when fetching encoding-variant web objects.
Amos
