Memory Utilization on HP- unix 11.23

[Chitta]

Hi All,
I have newly configured Oracle RAC wtih 2 nodes on HP UNIX 11.23 server with CFS.After i took the Ignite backup of whole server, i had noticed the memory utilization reached 94%. I am using Glance to monitor memory.But server is working fine.

Can any one tell me how glance software works on memory?

regs,
Chitta

[Duncan Edmonstone]

Umm, well it measures utilization of your systems memory... difficult to know what more to say unless you have a more specific question...

Looking at the memory bar at the top of glance you can see how much memory is being used by the kernel (between the S chars), user space (between the U chars) and buffer cache (between the B chars).

94% memory isn't necessarily bad - its good yo use that expensive memory you paid for. Are you completely sure it 'jumped up' after running Ignite - that seems unlikely to me, that said ypu can check which procs are using the most memory by running:

UNIX95= ps -eo vsz,comm | sed 1d | sort -rn | more

biggest memory users appear at the top of the list.

HTH

Duncan

[Chitta]

Thanks For quick reply Duncan,

i had allocated 2 gb for oracle process and 10% System buffer cache (dbc_max_pct parameter).even then the user memory shows 3.4 gb of utilization.
here are the glance report

Total VM : 3.4gb Sys Mem : 1.6gb User Mem: 3.4gb Phys Mem : 8.0gb
Active VM: 2.6gb Buf Cache: 816mb Free Mem: 2.1gb FileCache: na
MemFS Blk Cnt: na MemFS Swp Cnt: na

However,when i down the oracle, then memory comes down to 64%.So i think no problem with oracle.Now i am doubted about Glance functionality.



Regs,
Chitta.

[Duncan Edmonstone · Accepted solution]

Chitta,

WHat are you not understanding? You allocate 2GB of memory presumably for the Oracle SGA, which is shared memory. The other 1.4GB of user memory is no doubt for the Oracle shadow processes/PGA plus various daemons and processes that are part of a standard HP-UX install.

If your memory utilisation goes down by 30% after stopping Oracle, then you know that Oracle uses 30% of your 8GB memory, or approx 2.4GB. So at least 400MB is PGA and shadow processes.

HTH

Duncan

[Aashique]

Hi,
Glance output sometimes not comes accurate.
Use the "kmeminfo" script which will tell you per user basis memory and cpu utilization in details.

Thanks & Regards

Aashique

"Attachment removed as per ITRC GuideLines"

[Duncan Edmonstone]

Aashique,

kmeminfo is not supposed to be a customer tool - it generally does no harm, but isn't officially 'supported' (at least that I am aware of), and you really shouldn't be distributing it on public forums.

HTH

Duncan

[Aashique]

Hi Duncan,

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=972856

i got it from forum. i didnt know that cannot share it with anybody.

Thanks

[Rasheed Tamton]

Hi Chitta,

You have got 2.1GB free mem. That is more than enough not to worry about and more than 25% percent of your Phys Mem.

Total VM : 3.4gb Sys Mem : 1.6gb User Mem: 3.4gb Phys Mem : 8.0gb
Active VM: 2.6gb Buf Cache: 816mb Free Mem: 2.1gb FileCache: na


Rgds.,
Rasheed Tamton.

[Chitta]

Hi All,
Thanks for reply,
whenever i stop the oracle usermemory goes down to 1.4 gb from 3.4gb and no of user is one, Obvisouley root.

i forget mentioned the key point. The server is in Idle. i just configured Oracle and monitoring.

Now i have 2.1 gb of free memory, if the transcation starts, each oracle user should minimun took 10 to 15 mb atleast, min of 200 users logged into server, then it became more critical.

my problem is even then i stop the oracle and other process such as cluster packages.. the user memory shows 1.4 gb, apart from sys mem:1.6gb.


regs,
Chitta.

[Don Morris_1]

I'm not sure where the problem comes in.

With Oracle running, User memory is 3.4Gb. Stop it, and it drops to 1.4Gb. Ok... so Oracle consumes around 2Gb, likely mostly in the SGA. No surprises there.

Idle state is around 2.1Gb free, ~800Mb in buffer cache, 1.6Gb System, 3.4Gb User (primarily Oracle). You're worried about the 10Mb to 15Mb * 200 user potential cost (or around 2 to 3Gb variant load).

Well, the first important thing to note is that if you aren't under pressure -- free memory is simply memory the system hasn't found a use for. HP-UX is fairly aggressive at caching for performance. Your buffer cache is likely limited (the dbc_min_pct/dbc_max_pct tunables), but we can't say from Glance [or even from the very-much-not-supported kmeminfo without a good deal of work] exactly how much of the memory used as Sys right now is actually just cached in case the kernel needs it again. Once pressure begins to occur (free memory drops to around 1.5% of total memory), the system will first work to flush out those caches [gradually] as needed. So my recommendation would be, try putting the load on the box and see how it performs. What you don't want to see is swap outs in vmstat/Glance [processes being deactivated]. Some page outs will occur, and this is normal (file backed pages may be pushed out) -- but really, you just want to see Sys come down into your real steady state load.

[Sajjad Sahir]

Dear Chitta
u memory configuration is perfectly ok
but dear u oracle process is consuming more more space, so u have to check with oracle team what are the process consuming more space
thanks and regards

Sajjad Sahir

[Chitta]

YES, But i had raised SR with Oracle ,anthey concluded d there was no discrpancy on oracle process.

Oracle advised to check with Glance utility functionality.

But Still memory is high.where is the problem? still we cannot understand,

Regs,
Chitta

[whiteknight]

Chitta,

There are a few performance patches you
should install in your system.


PHKL_37653s 700_800 11.23 mmap(2),mlock[all](2),madvise(2),msync(2)

PHKL_36669 s700_800 11.23 mlock performance fix

PHKL_37000 s700_800 11.23 Dynamic Buffer Cache patch



Hope this help.

WK

don't forget to assign points

[Yogeeraj_1]

hi Chitta,

Now you have the confirmation from Oracle that there is no issue on their side. You must installed the latest patches as mentioned by WK above.

Assuming that you are running Oracle 10g, i would also suggest that you monitor your database using the OEM database/grid control application during the interval where you notice the memory utilisation reaching > 90%

you may get some clues there also.

hope to hear from you soon.

kind regards
yogeeraj

[Chitta]

Hi,
Thanks for update,

The first patch is already been in the OS, So pls suggest rest of two patches needs to be applied? and is that patch have any adhoc patches? or any dependencies?


regs,
Chitta.

[Yogeeraj_1]

hi Chitta,

>The first patch is already been in the OS, So pls suggest rest of two patches needs to be applied? and is that patch have any adhoc patches? or any dependencies?


You can get more information about the rest of the two patches from the ITRC Patch Database which is available at: http://www12.itrc.hp.com/service/patch/crumbRequest.do?crumb=patch.breadcrumb.main&BC=&admit=109447627+1209474276201+28353475

Also, when downloading these patches dependencies are automatically detected and you will be informed accordingly. You can then also select the dependencies to be downloaded.


hope this helps!
kind regards
yogeeraj