RE: [squid-users] Squid-2.5.PRE13 released, 2.5.STABLE1 planned

From: Pawel Worach <pawel.worach@dont-contact.us>
Date: Thu, 19 Sep 2002 22:44:52 +0200

Something must be really wrong with squid 2.5's memory allocation..

Snip from squid.conf
cache_dir diskd /cache1 8192 8 256 Q1=72 Q2=64
cache_dir diskd /cache2 8192 8 256 Q1=72 Q2=64
cache_dir diskd /cache3 8192 8 256 Q1=72 Q2=64
cache_dir diskd /cache4 8192 8 256 Q1=72 Q2=64
cache_dir diskd /cache5 8192 8 256 Q1=72 Q2=64
store_dir_select_algorithm round-robin
cache_mem 256 MB
memory_pools off

The os is:
FreeBSD proxya 4.6.2-RELEASE FreeBSD 4.6.2-RELEASE #0: Thu Aug 22
14:16:20 GMT 2002

MAXDSIZ and MAXSSIZ are both set to 1G which is as much memory as the
machine has.
After about 48 hours squid dies with this message:
FATAL: xcalloc: Unable to allocate 1 blocks of 4104 bytes!

top looks lite this after about 36 hours:
  PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
70488 nobody 2 0 836M 505M RUN 106:37 0.10% 0.10% squid

With squid2.4 cache_mem was set to 341M without any problems, is my
cache_mem still to large for 2.5?

Regards
Pawel

-----Original Message-----
From: Hegedus, Ervin [mailto:airween@amit.hu]
Sent: den 17 september 2002 08:16
To: Pawel Worach
Subject: Re: [squid-users] Squid-2.5.PRE13 released, 2.5.STABLE1 planned

Hello,
 
> We have been using Squid 2.5 in production as of pre10 on two proxy
> servers. Each proxy gets about 2500000 http requests / day and about
> 150000 icp.
>
> It has been stable as a rock until today when it died with this
> message:
> FATAL: xcalloc: Unable to allocate 1 blocks of 4104 bytes!
> Squid Cache (Version 2.5.PRE12-20020908): Terminated abnormally.
> CPU Usage: 9192.902 seconds = 5885.603 user + 3307.299 sys
> Maximum Resident Size: 747756 KB
> Page faults with physical i/o: 6896
> 2002/09/16 05:25:52| storeDirWriteCleanLogs: Starting...
> 2002/09/16 05:25:55| Starting Squid Cache version 2.5.PRE12-20020908
for
> i386-unknown-freebsd4.6.2...
>
> The box has 1Gb memory and cache_mem is 341Mb, squid did not leave a
> core file.

there is many way to resolve this:
- decrease your cache_mem, for ex: 341 MB -> 256 MB
- reconfigure your kernel, and recompile it, but look this
  kernel parameters:
  options MAXDSIZ="(256*1024*1024)"
  options MAXSSIZ="(256*1024*1024)"
  options DFLDSIZ="(256*1024*1024)"

  chenge the '256' to 1024, and it may good for you...

but first, read the comments in LINT file.

> Any ideas?
>
> Regards
> Pawel Worach
> Postgirot Bank / Nordea IT

greetings:
a.
Received on Thu Sep 19 2002 - 14:44:58 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:10:21 MST