Re: Memory use

If you have any questions, please feel free to contact me.

Tim Hearne
CAST Application DBA
Principal DBA

Centralized Development
AT&T Services, Inc.
e-mail:  timothy.hearne@att.com
http://intranet.att.com/its/cdtworx/content.cfm/home/

Agile Bronze Certified

AT&T CAST team mailbox:  g01494@att.com

PROPRIETARY INFORMATION
The Information contained herein is for use only by authorized employees of AT&T, and authorized

Affiliates of AT&T, and is not for general distribution within or outside the respective companies

OS Version:  Red Hat Enterprise Linux Server release 5.10 (Tikanga) (Linux xxxxxxx 2.6.18-371.11.1.el5 #1 SMP Mon Jun 30 04:51:39 EDT 2014 x86_64 x86_64 x86_64 GNU/Linux)

When I use top, I have recently observed memory usage of up to 190 GB but I am being told by our Server support team that it never uses more than about 12 GB of computational memory (including memory used by a small Oracle instance).  I have read on-line that because PostgreSQL uses shared memory that top (and pmap) tends to overstate the actual memory in use.

top - 09:47:30 up 330 days, 23:12,  1 user,  load average: 2.79, 3.37, 3.14

Tasks: 1525 total,   3 running, 1520 sleeping,   0 stopped,   2 zombie

Cpu(s):  3.0%us,  0.3%sy,  0.0%ni, 96.4%id,  0.4%wa,  0.0%hi,  0.0%si,  0.0%st

Mem:  264114472k total, 193001540k used, 71112932k free,   497484k buffers

Swap: 16779852k total,   697304k used, 16082548k free, 179534676k cached

Using the information found in Table 17-2. PostgreSQL Shared Memory Usage (http://www.postgresql.org/docs/9.2/interactive/kernel-resources.html), I have calculated the following for the instances:

8.4.1 instance:

9.2.10 instance:

--------------------

The observed values above for Shared disk buffers and WAL Buffers were captured using the “Current Value” in the Backend Configuration Editor of the pgAdmin III (v1.18) tool.  They are a point in time value. 

I question the calculation for Shared disk buffers as it is way too high.  I saw another post stating the calculation could be off by the block size.  If that is the case, the displayed values above would be divided by 1024.  In the postgresql.conf files, the shared_buffer values were entered as 800MB and 2048MB.

The small Oracle mentioned above has the following SGA:

Bottom line, does it seem reasonable based upon the above that the server could only be using around 12 GB of memory?  Is there another way to determine the actual memory being used by the instances?

Thank you,

Tim Hearne

PROPRIETARY INFORMATION
The Information contained herein is for use only by authorized employees of AT&T, and authorized