Re: Low values for cached size

On Fri, Sep 25, 2009 at 3:28 PM, Carlos Henrique Reimer
<carlos.reimer@opendb.com.br> wrote:
> Hi,
>
> We're facing performance problems in a Linux box running CentOS release 5
> (Final) and PostgreSQL 8.2.4. I've done some basic checks in the
> configuration but everything looks fine to me. One weird behaviour I've
> found is the cached size showed by the
> "top" and "free" Linux commands:
>
> top - 08:32:17 up 3 days, 19:04,  1 user,  load average: 1.09, 1.07, 1.10
> Tasks: 173 total,   2 running, 170 sleeping,   0 stopped,   1 zombie
> Cpu(s):  9.5%us,  0.5%sy,  0.0%ni, 88.2%id,  1.7%wa,  0.0%hi,  0.0%si,
> 0.0%st
> Mem:   3631900k total,  3378056k used,   253844k free,    25488k buffers
> Swap:  4192956k total,      100k used,  4192856k free,  2356588k cached
>
> [postgres@server01 etc]$ free
>              total       used       free     shared    buffers     cached
> Mem:       3631900    3174804     457096          0      14280    2086184
> -/+ buffers/cache:    1074340    2557560
> Swap:      4192956        108    4192848
> [postgres@server01 etc]$
>
> Both commands show values ranging from 2GB to 2.3GB for the cached size and
> the server has 3.5GB RAM. I do usally see  cached values with sizes bearing
> the size of the RAM in other servers. It seams that something is consuming
> the RAM and not letting it free to be used as cache for Linux files, right?
> The shared_buffers (256MB) is not high and I can not see a reason for this.
> Initially I've thought the problem was
> because the system was running with runlevel 5, but now, it's running with
> runlevel 3 and even so the values for
> cached size does not change.
>
> Any suggestions or directions I could follow to discover the reason?

If you run top, then hit M, and post the first 20 or so rows after
what you have here I can take a guess.