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

From: Dedy Kurniawan <dedyk101_at_gmail.com>
Date: Sat, 30 Aug 2008 22:46:01 +0700

On Fri, Aug 29, 2008 at 7:09 AM, Chris Robertson <crobertson_at_gci.net> wrote:
> Dedy Kurniawan wrote:

> SNIP
>>
>> tcp_outgoing_address 10.5.9.106 typeofclient1
>> tcp_outgoing_address 10.5.9.107 typeofclient2
>>
>
> What does the acl typeofclient1 look like?

acl typeofclient1 src 10.5.8.105
acl typeofclient1 src 10.5.8.106
just labeling the client src addresses so that they could come out as
another ip which i could manipulate (ex: routing, QOS)

>
>
> SNIP
>
> Lots of good info, but what OS? What kernel?

kernel ubuntu 7.10, kernel 2.6.22-14-server

>> 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
>>

> Why do you do this?
because i read about memory management using malloc so i try to
configure squid using it

>> 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
>>

> This might be effected by the headers returned with the object. Using
> "no_cache allow all" is not recommended with the default refresh_pattern,
> due to the likelihood of caching dynamically generated content.

okay, so you suggest me to set the default,
no_cache deny QUERY

>> 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
>> :((
>>

> Perhaps it's related to the malloc libraries you are using... The only
> information I could find about libgnumalloc indicates that it's pretty
> outdated, which indicates that the underlying OS is... Aging.

okay, do i have to re-install squid? i will if i have too :)

>> 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?
>>

> I don't know of any reason not to. But for limiting bandwidth, delay_pools
> might be a better choice.

i use squid with together with Mikrotik

Internet ---- Mikrotik --- Squid --- Client

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

> Update your Squid version, and perhaps your OS.

which version of squid you suggest me?

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

-- 
~dedy
Received on Sat Aug 30 2008 - 15:46:05 MDT

This archive was generated by hypermail 2.2.0 : Sun Aug 31 2008 - 12:00:04 MDT