Memory Problem

[Sanjiv Sharma_1]

Dear Friends,
I am facing the following problem in my HP Server.

Operating System : HP Ux 10.20 .
Server : HP 9000 K Class K580
Memory : 4GB RAM .
Application : I2O

Available memory shown by the system is only 2GB RAM whereas total RAM in
the server is 4GB .If the " dmesg " o/p is seen , we are able to notice the
following parameters :

Total Physical Memory : 4GB
Lockable Memory : 1.8GB
Available Memory : 2GB .

It is to be noticed why only 2GB RAM is being told available by the system.

Pls. help me to resolve this.

Thanks & regards ,
Sanjiv Sharma

[James R. Ferguson]

Hi:

Hopefully this will help you you understand better:

/usr/share/doc/mem_mgt.txt

...JRF...

[Stefan Farrelly]

available memory is the total memory left free after the kernel has loaded into memory and setup all its tables, buffers and buffer cache. Have you got the buffercache set to a very high value ? You can check with glance or gpm, or if you dont have these tools do;
echo bufpages | adb /stand/vmunix /dev/kmem
The resulting number of bufpages is in 4k pages, so times it by 4 for how many MB being used by buffercache. You can restrict this number by using the kernel parameter dbc_max_pct and db_min_pct.

[James R. Ferguson]

Hi:

# sysdef

will also show you your static values for dbc_min_pct, dbc_max_pct & bufpages.

...JRF...

[Andy Monks]

Looks like you've got some very large kernel structures. What values you do have for all the semaphore kernel parameters? sem*

[James R. Ferguson]

Hi:

Also forgot to add that around 10.10 (I think) the default settings for dbc_max_pct was 50% -- rather high. If you have glance, you can use its IO metrics to contrast physical vs. logical IO of your disks to gain an idea of how much the buffer cache is benefiting you or how much memory you are wasting. Glance also shows the buffer cache as "B - B" in its memory utilization bar graph at the top of each screen.

...JRF...

[Anthony deRito]

Sanjiv, the values you are seeing is simply a result of one of your kernel parameters called unlockable_mem. I am guessing that this value is set to 0. This is the default value. If this is the case, then the kernel will chose an appropriate value at boot time.

The value for lockable memory seems high but this is normal. This simply means that it is the highest amount of physical memory that users can lock down at once. Typically, users may want to lock memory into physical memory for performance reasons. They do this by system calls like plock().

If your kernel parameter unlockable_mem is not set to 0 and/or if you suspect memory problems for some other reason, then you need to concentrate on performance issue.

Tony

[John Palmer]

My guess is that the server has a very large fixed buffer cache.

Either 'bufpages' is set to a large value or dbc_max_pct/dbc_min_pct are set to something like 50%.

[CHRIS_ANORUO]

Hi Sanjiv,

You don't have to worry about this. Use swapinfo -tm (under type you have memory if swapon_mem has been set to 1 (one) in kernel parameters)
ie swapinfo -tm
memory=75%(Physical memory)=3Gb., if this is true, then you son't have a problem.
Also do cat /var/adm/syslog/syslog.log|grep Phy(Physical:4Gb)
From kernel configuration, set dbc_max_pct=25 and dbc_min_pct=5