Memory usage is only 14% still swapping is happening. why?

[Nafez ALNajjar]

Hi,
vhand daemon will give an indication of how busy your system with swapping try :
#ps -ef |grep vhand

also find the attached document wich may help in troubleshooting java memory related issues.

[chris huys_4]

>Chris: The standard practice on hp-ux, is
> still to create as much device swapspace as
> there is physical memory.

> Dennis: This practice may be wasteful for
> large memory systems.
Not as long as 4Gbyte of memory on a hp-ux itanium system costs 4 times the price of a 146Gbyte internal disk. ;)

Greetz,
Chris

[ManojK_1]

Hi,

I have raised the issue with HP and the application vendor and they were working on it.

If the issue got solved i will update the details in the forum.


Manoj K

[SoorajCleris]

Hi Manoj,

Your OS version is 11.31 and if you are talking about vhand deamon activities,this is as an expected behaviour.

Vhand handles aging and stealing pages in the filecache.

When you see this in 11.31 , don't worry untill and unless system has memory pressure.


Regards
Sooraj

[Mancboy]

with java bases systems it is always better to have more swap than there is RAM.
Java will and does leak, we've seen java processes locked at 2Gb, take 15Gb.
HPUX reserves the memory in swap "just in case it needs to use it", if there isn't enough swap available to reserve, it won't schedule the job - I've seen this enough times after raising countless calls to HP on this.
Also, java will sap system memory - kmeminfo shows how poor HPUX is, if you have glance installed you can see the memory getting hived off to system over time.

HP still recommened twice the swap of memory, and even though I'm opposed to this waste of money, it has proven very useful in keeping the systems up and running.
So, to keep the system afloat, 64Gb swap, to guarantee no vhand / swap issues, 128Gb Swap.

Trust me, in the long Java run, it's worth it just to shut syslog up.

[Rita C Workman]

Did read every line above....so forgive me is this is covered.

But on that vmstat that shows page outs.....have you cleared the accumulators on vmstat to ensure you're seeing current issues, and not just old info.

man vmstat

vmstat -z

Then monitor for page-outs..

/Rita