[squid-users] xmalloc and out of memory errors in messages log

Hi,

I am currently having these messages in my
/var/log/messages:

Sep 20 00:09:58 blar out of memory [28008
Sep 20 00:09:59 blar squid[29539]: Squid Parent: child
process 28008 exited due to signal 6

Sep 20 01:39:21 blar (squid): xmalloc: Unable to
allocate 87380 bytes!
Sep 20 01:39:21 blar squid[29539]: Squid Parent: child
process 28435 exited due to signal 6

During peak hours, I'm getting about 8 xmalloc errors
per hour. out of memory are less frequent at about
twice a day. I've read the FAQ. Its states 2 possible
reasons for OS to have xmalloc error:
1) out of swap
2) data segment size reached

I am sure that the machine never ran out of swap,
using monitoring tools. And the below shows that I've
set the data segment size to unlimited.

[root@squid squid]# ulimit -a
core file size (blocks, -c) 0
data seg size (kbytes, -d) unlimited
file size (blocks, -f) unlimited
max locked memory (kbytes, -l) unlimited
max memory size (kbytes, -m) unlimited
open files (-n) 1024
pipe size (512 bytes, -p) 8
stack size (kbytes, -s) 8192
cpu time (seconds, -t) unlimited
max user processes (-u) 7168
virtual memory (kbytes, -v) unlimited

What else can I do for the xmalloc and out of memory
errors?

my caching disk partitions(reiserfs):

/dev/sdb1 34G 22G 12G 64% /cdata1
/dev/sdc1 34G 22G 12G 64% /cdata2
/dev/sdd1 34G 22G 12G 64% /cdata3

linux kernel 2.4.20-19.8

DELL 2.4 Xeon SP, 2G RAM, 2G swap,
2X34G mirror for OS, 3X34G volume for cache

About 120 req/s(non-peak) and 250 req/s(peak).

Squid Cache: Version 2.5.STABLE3(patched deny_info
err)
configure options: --prefix=/usr/local/squid
--enable-dlmalloc --enable-removal-policies=heap
--enable-storeio=aufs --enable-underscores
--enable-async-io=50 --enable-snmp

squid.conf(important sections)
==========
cache_mem 64 MB
cache_swap_low 92
# cache_swap_high 95
maximum_object_size 1024 KB
maximum_object_size_in_memory 100 KB
cache_replacement_policy heap GDSF
memory_replacement_policy heap GDSF
cache_dir aufs /cdata1 23800 56 256
cache_dir aufs /cdata2 23800 56 256
cache_dir aufs /cdata3 23800 56 256
quick_abort_min -1 KB
half_closed_clients off
memory_pools off

One more thing: I've another squid(identical in
everything except cache space) with 16G X 3 of caching
space and they exhibit the same freq of errors.

Pls help. Let me know if I've left out anything.
Thanks.

--
Wolf
__________________________________________________
Do You Yahoo!?
Play now and stand a chance to win cash prizes! 
http://yahoo.com.sg/millionaire