Re: [squid-users] A Different Squid Load Issue?

From: Chris Robertson <[email protected]>
Date: Thu, 08 Feb 2007 17:33:49 -0900

Chris Nighswonger wrote:
> Hi all,
> I have a issue where when the client browser requests a page, it
> (the client) appears to hang and just wait without loading the page.
> After several 'refresh' attempts, the page will load. At other times
> the client behaves as if the page has loaded, but displays nothing
> (blank screen).
> Over the past two days I have watched the squid process via 'top'
> and notice that when this phenomena occurs the squid process grabs
> %100 of the cpu resources and the 'top' screen freezes until squid
> backs off. Most of the time squid uses far less than %30 of the cpu.
> I have followed several of the threads about squid under high load
> and these do not seem to apply here.
> Additional info that may help:
>
> Squid Version 2.6.STABLE5
> Hardware: Dual PII 400MHz, 192MB ram, 40GB LVM (2 odd sized SCSI drives)
> OS: Fedora 6
> The system also runs DansGuardian, a firewall, and a caching-only
> Bind9 DNS
> There are only 40-50 users avg connected simultaneously. Some use VOIP
> applications (ie. Skype, etc.)
>
> Some pertinent lines from squid.conf
>
> store_avg_object_size 14 KB #Based on cachmgr.cgi avg
> cache_dir ufs /var/spool/squid 20480 65 256
>
> Does anything here stand out as a potential issue?

The cache_dir type. Look into replacing it with aufs, to "avoid
blocking the main Squid process on disk-I/O." This might require a
recompile of Squid (if you are not using the Fedora 6 RPM), as ufs is
the only one built by default. Another pitfall is the small amount of
memory the server has, and the large size of the disk cache. See
http://wiki.squid-cache.org/SquidFaq/SquidMemory#head-09818ad4cb8a1dfea1f51688c41bdf4b79a69991
for more details.

>
> Thanks
> Chris
>

Chris
Received on Thu Feb 08 2007 - 19:34:00 MST

This archive was generated by hypermail pre-2.1.9 : Thu Mar 01 2007 - 12:00:01 MST