At 21:55 09/06/98 -0700, you wrote:
>That's right
>
>we got very big sr on the vmstat result, i'll shrink the cache_mem value
to 128,
>and hope that work :)
>btw, what about the dlmalloc is there any idea ? i got problem to compile
>that....(or should i use that ?), and about the NOVM version i got the same
>problem the daemon crash for a little period.
If I were you I would either double the RAM in the machine, or half the disk
space allocated for the cache. We have a squid cache running with 12 GB
dedicated for cache and 512MB of RAM. It is Solaris x86 2.5.1 with fairly
recent patch cluster installed. The configs for cache_mem and cache_swap
are as follows:
cache_mem 220
cache_swap 10000
and the memory consumption according to top is the following:
PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
14768 root 34 -15 398M 309M sleep 145.6H 10.78% 10.28% squid
From my own experience, decreasing the cache_swap variable alone would
cause a considerable decrease in memory consumption.
Marco Aurelio S. Lopes
Internet Consultant - Unisys Network
Unisys Brasil Ltda.
Sao Paulo - SP - Brazil
Email: marco@unisys.com.br
Fax: +55-11-867-9005
Received on Tue Jun 09 1998 - 09:56:17 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:40:40 MST