Glance: Mem Util averages 96%

[Wilmar Ricio]

I'm using HP10.20, 640mb Phys. Memory w/ 60 users. Using glance, my Mem Util stays around 96%. I reboot system every month end, goes down to about 90% then up to 96% to 98% by month end. Is this something to worry.
Also, under the RSS column, some users consume more memory. Is there a fix for this.

Thanks.

[James R. Ferguson]

Hi:

The first thing that comes to mind is your buffer cache. What are your values for DBC_MIN_PCT and DBC_MAX_PCT ? You can do 'sysdef' to show them. If DBC_MAX_PCT is 50%, then up to 50% of your memory can be used for file buffers.

...JRF...

[John Palmer · Accepted solution]

Hi Wilmar,

HP-UX will try to make best use of the memory that you've got by using as much of it as it can.

What probably changes on your server is the size of the buffer cache - providing that you have configured the kernel with the parameters dbc_max_pct (default 50) and dbc_min_pct (default 5). These defaults allow the buffer cache to expand to 50% of your available memory (320Mb) when you are not using it and shrink to 5% (32Mb) when you are. You can check these settings with glance on the 't' screen (page 2).

I tend to judge how much memory is in use not by the used or free percentages but by the size of the buffer cache.

The RSS column (Resident Set Size) tells you how much of a process is actually paged into main memory as opposed to it's actual size (VSS - Virtual Set Size).

The RSS size of a process tends to vary as parts of it get paged in and out, I don't know what you want a fix for, perhaps you could give some more information.

Regards,
John

[Dennis Kennedy]

whats up!

It sounds like you are woried about a memory leak in one of your processes. run "ps -efl | more" or "ps -efl > afile" then vi afile and look under the "S" column (state).
0 nonexistent
S sleeping
W waiting
R running
I intermediate
Z terminated
T stopped
X growing

if you see an X then the process has a memory leak and is growing in size.

Hope this helps
later,
D