Henrik Nordstrom wrote:
>mån 2006-03-27 klockan 17:31 +0200 skrev Mark Elsen:
>
>
>>>Mar 27 17:14:11 w3cache1 Out of memory: Killed process 28301 (squid).
>>>
>>>
>>Seems like SQUID ran out of memory.
>>
>>
>
>Or more precisely that your server ran completely out of memory, and as
>a last resort it in a panic killed the most memory consuming process to
>try to free up memory allowing the OS to survive.
>
>Regards
>Henrik
>
>
Server have 4GB of ram
P4 3.2GHz
8 scsi disks
squid-conf
###
http_port 8080
icp_port 3130
hierarchy_stoplist cgi-bin .cgi .asp ?
acl QUERY urlpath_regex cgi-bin \? \.asp \.cgi
no_cache deny QUERY
hosts_file none
cache_mem 64 MB
cache_swap_low 80
cache_swap_high 100
#debug_options 43 1
maximum_object_size 32768 KB
minimum_object_size 0 KB
maximum_object_size_in_memory 96 KB
ipcache_size 65535
fqdncache_size 65535
cache_replacement_policy heap LFUDA
memory_replacement_policy heap GDSF
###
And 8 cache dirs with coss or aufs with max space for cache 17GB each.
With coss of aufs i have the same problems.
squid is running for 3-4 hours and crash.
This proxy is transparent for 4500users....
Cpu load without epool patch 100% always
Cpu load with epool patch 20% max
cat /proc/net/sockstat | grep TCP
TCP: inuse 12845
Received on Thu Mar 30 2006 - 09:30:54 MST
This archive was generated by hypermail pre-2.1.9 : Sat Apr 01 2006 - 12:00:05 MST