[squid-users] squid-2.5.STABLE3 based server hangs periodically

Hi !

I'm running subject with RedHat Linux 7.3, w kernel 2.4.21. Server hardware
is IBM x342 with 1,13GHz P3 processor, 1GB RAM & ServeRaid controller
running RAID0 stripeset. Filesystem on cache partition are reiserfs and
cache_dir size 40MB. Average traffic are about 20Mbit/s and about 150-200
request per second. Problem is, that system hangs completely after couple
of days. There are no any errors or any abnormal things on the log files,
everything looks ok, but.... I tried to use completely another server
hardware based on IBM x340, but problem are the same. Squid are self
compiled with increased filedescriptors limit. Configure options are:

./configure --prefix=/usr --sysconfdir=/etc/squid --mandir=/usr/share/man -
-localstatedir=/var --enable-snmp --enable-delay-pools --enable-linux-netfil
ter --enable-default-err-language=Estonian --enable-async-io --enable-storei
o=diskd,aufs

Also i tried different io modules, diskd & aufs, but there was no luck,
except that sometimes diskd crashed, and squid was restarted. Any ideas ?
kernel problems ? or heavy traffic ? Typically crash or hang happens when
cache is full, traffic high, and processor usage about 90%, but last hang
was at midnight when traffic was only 2Mbit/s and about 20 request per
second. There are also one bad thing more. When system hangs or squid
crashes, then cache storage becomes unusable, squid thinks that the storage
size is over limit and makes unnessesary deletion of storage files. I played
also with different config options, like cache_mem, cache_dir size,
maximum_object_size and others, no luck.
I'm running about 25 different servers with same hardware, operating system
& kernel, but those things happens only with squid based proxy servers. OK,
there are no such kind of traffic on the other servers, except one ftp
server, where average traffic is about 250Mbit/s.

Regards.
Markos
Received on Sat Aug 09 2003 - 14:01:02 MDT