[squid-users] help... anyone :-)

Hi all,

1st of all, i'm sorry before, for posting my current squid condition,
from how i install to current squid config. Currently i'm having a
problem with my squid, related to cache. I hope all of my friend here
could help me. And i hope this thread useful to anyone having trouble
like mine :-)

ok here we go guys, this is my config:

-----start config snippet--------------------------------------

cache_mem 6 MB
cache_swap_low 98
cache_swap_high 99

maximum_object_size 16 MB
#minimum_object_size 0 KB
maximum_object_size_in_memory 64 KB

cache_replacement_policy heap GDSF
memory_replacement_policy heap GDSF

cache_dir aufs /cache1 6000 14 256
cache_dir coss /cache2 6000 block-size=512 max-size=131072
#cache_dir aufs /caches1 6000 14 256 <--- sometimes i used this 2
lines to see the differences from the upper 2 lines
#cache_dir aufs /caches2 6000 14 256 <

cache_access_log none
cache_log none
cache_store_log none
cache_swap_log /var/log/squid/%s

refresh_pattern ^ftp: 10080 95% 241920 reload-into-ims override-lastmod
refresh_pattern ^gopher: 1440 0% 1440
refresh_pattern . 180 95% 120960 reload-into-ims override-lastmod

ipcache_size 4096
ipcache_low 98
ipcache_high 99
fqdncache_size 4096

quick_abort_min 0 KB
quick_abort_max 0 KB
quick_abort_pct 98

reload_into_ims on
pipeline_prefetch on
ie_refresh off
vary_ignore_expire on

digest_generation on
digest_bits_per_entry 10
digest_rebuild_period 30 minute
digest_rewrite_period 30 minute
digest_swapout_chunk_size 6000 bytes
store_dir_select_algorithm round-robin
server_persistent_connections off

tcp_outgoing_address 10.5.9.106 typeofclient1
tcp_outgoing_address 10.5.9.107 typeofclient2

no_cache allow all

----end config snippet--------------------------------------

model name : AMD Sempron(tm) Processor LE-1100
stepping : 1
cpu MHz : 1908.810
cache size : 256 KB
memory : 3Gb

current client : apprx 200 client (divided in 3 cluster regarding
their request of service)
with current setting, squid respons in a short time, for hit and miss.

squidbox linux misc conf :
in /etc/rc.local ---- > ulimit -n 8192;/usr/local/squid/sbin/squid -sY
-f /etc/squid/squid.conf;

when i compile squid, this is what i do

root_at_proxy malloc# vi Makefile
    uncomment -> #CPPFLAGS = -DUSG
root_at_proxy malloc# make
root_at_proxy malloc# cp libmalloc.a /usr/lib/libgnumalloc.a
root_at_proxy malloc# cp malloc.h /usr/include/gnumalloc.h

root_at_proxy malloc# vi /etc/security/limits.conf
* Soft nofile 8192
* Hard nofile 8192

root_at_proxy malloc# vi /etc/sysctl.conf
    net.ipv4.ip_local_port_range = 1024 32768

root_at_proxy malloc# cd ../squid_2.5.STABLE14

------the question -------------

my question is :

1. i wanna cache big files, example from antivirus update, but when i
check the log, file like u7avi1363a.bin (6MB size) is always missing
    i even desperately try --------> no_cache allow all

2. i don't know, maybe only my feeling, but why is when the memory is
runout, i get slow responses from squid. I check it like this

free -m
root_at_proxy:~# free -m
             total used free shared buffers cached
Mem: 2978 2891 87 0 91 2565
-/+ buffers/cache: 234 2744
Swap: 368 0 368

i read hendrik's reply in older thread before, he said that look at
buffers/cache not the Mem , but why the responses is different after i
reboot my squid, which will give me free in Mem:
with this kind of behavior, my squid uptime is always less than 12 hours :((

3. currently i'm running about 15 tcp_outgoing_address so every client
come out from squid with their ip alias address, so that i can limit
their bandwidth. is it okay to run so many ip alias in linux, and
running many tcp_outgoing_address in squid?

4. any suggestion for my config? i mean to speed up responses when
opening a web and to deliver it to clients

thanks for reading my question, and many thanks for the upcoming answers ;-)

-- 
~dedy
--
~dedy