[squid-users] Corrupted Downloads

> I'm getting corrupted downloads from Squid 2.5Stable1 running on Red Hat Linux 7.2.
> Large files are occasionally corrupted. The byte count of the file is correct, but somewhere
> in the corrupted file I'll find another download's HTTP header and data.
>
> Brett Lymn reported the same problem
> (see http://www.squid-cache.org/mail-archive/squid-users/200110/1083.html)
> almost a year ago.
> In May Alan Flavell reported the problem, too
> (see http://www.squid-cache.org/mail-archive/squid-users/200205/0349.html)
>
> Phil Oester has the most detailed analysis of the problem to date
> (see http://www.squid-cache.org/mail-archive/squid-users/200210/0137.html and
> http://www.squid-cache.org/mail-archive/squid-users/200210/0311.html)
>
> I built squid 2.5Stable1 on RedHat Linux 7.2 with the following config options:
> configure --enable-cache-digests --enable-underscores --with-aio-threads=30 --with-pthreads
                --enable-storeio=aufs,ufs --enable-removal-policies=heap

> I used gcc version 2.96 20000731 (Red Hat Linux 7.2 2.96-108.7.2) to build squid. The
> machines running Squid are Dell 2550 servers, each with dual 1 Ghz CPUs and 1 GB
> of RAM. They each have 15 GB of cache disk. There are three of them configured as
> siblings and I use round-robin DNS records to balance clients between them. Each
> squid cache handles about 60 requests a second.
>
> Here are some of the relevent options in the squid configuration file:
> Cache_peer proxy1 sibling 3128 3130 proxy-only
> Cache_peer proxy2 sibling 3128 3130 proxy-only
> Cache_mem 32 MB
> Cache_swap_low 95
> Cache_swap_high 97
> Maximum_object_size 128 MB
> Maximum_object_size_in_memory 24 KB
> Cache_replacement_policy heap LFUDA
> Memory_replacement_policy heap LFUDA
> Cache_dir aufs /opt/squid/var/cache 15000 53 251
> Cache_store_log none
> Redirect_program /opt/adzap/wrapzap
> Redirect_children 10
> Ident_lookup_access allow all
> Client_db off
> Redirector_bypass on
>
>
>
>
>
> Are there any Squid developers working on the problem?
>
>
>
> Mike Mitchell
> SAS Institute Inc. The power to know
> Mike.Mitchell@sas.com
> (919) 677-8000 x16793
>
Received on Wed Oct 23 2002 - 09:38:32 MDT