Re: [squid-users] Squid Accelertor Problem with Cache_Peer's

Henrik Nordstrom wrote:

>mån 2007-03-05 klockan 12:32 -0500 skrev Shane A. Froebel:
>
>
>
>>We at the battlestarwiki.org are using Squid in accelerator mode at the
>>moment to serve as the primary point for all web-connections. However,
>>everytime we have like a horde of users access the site, the server
>>itself goes slow. So I tried settings up a secondary cache server on our
>>SQL server (which only has SQL and no httpd). Now I read everything I
>>can and I can still not get everything to work correctly.
>>
>>
>
>I don't quite get how you want the relations between the two.. should
>users connect to one or both?
>
>From your configs it looks like users always connect to the Squid on the
>same server as the web server. Under these conditions it won't generally
>help to have this Squid bounce the requests to yet another Squid.
>
>What you can do to improve performance is to have another server also
>accept requests from the users, and forward them to the web server.
>
>But most importantly you need to diagnose why things gets slow. So I'll
>follow up with a bit of questions focusing on your first server for now:
>
>* Is there any swap activity under load?
>* How is the Squid cache configured?
>* What is the primary load condition? CPU? I/O? Memory?
>
>
>Regards
>Henrik
>
>
Apollo and Athena are two different machines.

1)
22:14:17 up 14 days, 28 min, 1 user, load average: 34.79, 28.25, 27.54
-- Been like this since after the last episode aired.
             total used free shared buffers cached
Mem: 2067852 832508 1235344 0 104556 264192
-/+ buffers/cache: 463760 1604092
Swap: 2040212 355476 1684736
2) Right now Apollo uses the default dir /usr/local/squid/var/cache
(5gb) and since we don't use much "data" space for files because of the
mediawiki setup, a 10gb space under /home/squid_cache which is 10gb --
running AS root because it runs on port 80. Apache 1.3.3 runs on 80, but
bind to 127.0.0.1 IP
3) Each machine has a Intel \ 2.4 Ghz \ P4 Celeron, and 2gb of ram in
apollo and 1gb on athena. Both 100mb uplinks. Seagate \ 80GB:IDE:7200RPM
Barracuda \ ST380011a - Redhat Enterprise Linux - OS ES.

Apollo has cPanel.

Talking with my friend, we are thinking about just changing to a more
powerfull machine that can support muli-cpus. So any help would be great.

On athena, we have MySQL bind to one IP, and I have, in-thery,
squid.battlestarwiki.org on another IP.

-- Shane
Received on Mon Mar 05 2007 - 21:30:28 MST