High Memory Utilization Reported by Glance

[HP_Sammy]

Hello All,

We have a server which is having a high memory utilization.It is still not in production and very less users connected but still its using lots of memory space i have investigated the issue and it seems oracle db is using the max memory and for the same i have attached my observation in a notepad file could you guys please look into the notepad file and help me fix the issue or provide your suggestion.

Thanks all in advance.

Regards
Coolsami

[HP_Sammy]

Guys please open using open with in notepad it will work.

Regards
Coolsami

[Dennis Handly]

You need to change swapmem_on to 1.
Your dbc_max_pct is 9 & dbc_min_pct is 6.

Why do you think memory use is a problem? Is the server working fine?

[HP_Sammy]

hey Dennis thanx for the reply please check the glance output in the notepad its almost reaching 90% i think server is not having issue but still glance op is a concern and oracle is eating up lot of memory.

Regards
Coolsami

[Ganesan R]

Hi,

I could see that you are having 4GB of physical memory and you have oracle on the server. 4GB is not sufficient for oracle servers.

You said only oracle processes are occupying more memory. So it is clear that user process consumes all the memory. Either you need to work with oracle to find if those process really need that much memory or not. If yes increase the memory.

[HP_Sammy]

hey Ganesan thnx for the reply but question here why do u say that for oracle 4gb memory is not enough ??? is it documented some where and i have already involved DBA team to look into the issue they working on it in the mean while wanted to know if itrc hp ux guys can analyse issue and provide der suggestion on the issue and for the same i have attachhed the notepad and also adding more memory is not an option bcoz the client has not approved it.

Thanks
Coolsami

Note : i will provide 1 point for each reply from all hp ux itrc guys.

[Ganesan R]

Hi,

It is my personal experience that oracle db server needs more memory depends on how many instances and the user sessions. If you need document probably you can get it from oracle on the memory needs.

There are worthful tools available from HP to get into deep on which process consumes how much physical memory. one is kmeminfo which will provide memory usage on each and every procoess.

Based on that you can decide how much memory oracle process are taking.

I really like your quote:
"Note : i will provide 1 point for each reply from all hp ux itrc guys"

[TTr]

> Note : i will provide 1 point for each reply from all hp ux itrc guys.

Why? Their answer is not under an HP maintenance agreement or a call that you placed with HP support at ITRC. This is the free HP forums. Check out the forum etiquette at http://forums13.itrc.hp.com/service/forums/helptips.do?#overview

Now, oracle is using 1.2GB of memory in shared mem segments. Usually that's the size of the oracle SGA. To that you need to add all the memory that the running oracle processes use. Many oracle binaries are 200 - 300 MB in size and you have at least 20 oracle processes running. Shared pages do help with this but up to a point. Then you have the OS overhead. It appears that you are running CIM and/or SIM, if you don't need them you should turn them off along with the apache web server. And last you have the OS daemons which you did not list them all, buffer cache which is 360MB and the kernel itself.

In glance you should hit the "m" key to show the memory details.

I also think that 4GB is not enough for a database server these days. You have a rp8420 with 11.23, the HP-UX 10.20 days with minimal memory usage are gone.
Once you users start to hit the database your memory utilization will probably increase accordingly.

[Emil Velez]

of course. The dynamic buffer cache should allow all memory not used by processes to mostly be used by buffer cache.

The main question is

1. Are you ever getting low on memory where you are doing page outs. If you are doing any page outs you are low on memory (or DBC_max is too high)

2. Are your processes waiting on VM ? or can your processes run as fast as they can and never waiting on VM. (look at global wait states).

If you dont have a memory bottleneck look for other stuff.