topic Re: vhand eating up the CPU time ! in Operating System - HP-UX

vhand eating up the CPU time !

Deependra — Thu, 12 Aug 2004 05:27:10 GMT

Hi,

We are facing the performance in one of our system. The system used to run fine earlier but suddenly it becomes slow and then goes to not responding mode. I checked and found that vhand is eating up most of the CPU time.

I checked the swapifo and found that it is not even used 50% below is the output of swapinfo:
===========================================
# swapinfo -at
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 8192000 170504 8021496 2% 0 - 1 /dev/vg00/lvol2
reserve - 4313668 -4313668
memory 3158580 48408 3110172 2%
total 11350580 4532580 6818000 40% - 0 -
==============================================

The system is having 4 GB RAM, HP-UX version is 11.11 and the model is J6700.

Can anybody provide some suggestions about what could be the reason.

Thanks
Deependra

Re: vhand eating up the CPU time !

john kingsley — Thu, 12 Aug 2004 06:23:32 GMT

Sounds like you're running out of memory. Vhand is the process the move pages in and out memory.

Check to see what your kernel parameter "dbc_max_pct" & "dbc_min_pct" are set too. By default, dbc_max_pct is set to 50% which is way too high. Lowering these parameters will make more memory available. We typically set dbc_max_pct to 10% and dbc_min_pct to 5%. On systems with a lot of memory, we will set dbc_max_pct to as low as 5%.

Re: vhand eating up the CPU time !

Kent Ostby — Thu, 12 Aug 2004 06:29:52 GMT

Deependra --

There are some known problems with vhand on 11.11 eating up CPU.

The fix is in the vhand patch: PHKL_30796

Best regards,

Kent M. Ostby

Re: vhand eating up the CPU time !

Steve Steel — Thu, 12 Aug 2004 06:34:17 GMT

Hi

For the memory used you need to add the reserve which is swap held by processes in memory so while not used no longer available; This is so the process can swap if needed

1)Do you have
PHKL_28695
: s700_800 11.11 Cumulative VM, Psets, Preemption, PRM, MRG

This will help with these situations.

2)As advised low the dbc_max which will free memory

3)To check memory usage see

Memory Usage - â What is using all of the memoryâ ?

by:eric.herberholz@hp.com

Last modified: July 26, 2004

Latest version available at external ftp site:

ftp://eh:spear9@hprc.external.hp.com/memory.htm

Which gives a good explanation and some unofficial tools

Steve S

Re: vhand eating up the CPU time !

Deependra — Thu, 12 Aug 2004 09:36:23 GMT

Hi,

The dbc_max_pct and dbc_min_pct were 10 and 5 respectively and the patch PHKL_28695 was also installed. HP also suggested to install the patch PHKL_30796 along with PHKL_30904.

We have installed these patch and need to observ if the problem occurs again.

What I am wondering that the system was working fine earlier and suddenly it started behaving in this way. There is no message in the syslog regarding any hardware problem also !

Any clue on this !!

Thanks
Deependra

Re: vhand eating up the CPU time !

Patrick Wallek — Thu, 12 Aug 2004 09:52:11 GMT

This has almost nothing to do with hardware. This is almost certainly due to ALL of your RAM being used and the system has to start paging processes out to the swap space so new processes can start.

It may be that someone started a large process or a large query or something that ate up all of your RAM.

The only hardware that may help you in this case is RAM. As in install some more of it.

Re: vhand eating up the CPU time !

Deependra — Tue, 17 Aug 2004 02:35:14 GMT

Hi,

After applying the patches mentioned in my previous mail the frequency of system hanging is reduced but the problem is not resolved completely. Is there any solution other than increasing the RAM? We already have enough (4GB) RAM and it used to work fine earlier. We have tried to increase the swap space but it is not being used at all (see the output of swapinfo given below)
=====================================
# swapinfo -tm
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 8000 0 8000 0% 0 - 1 /dev/vg00/lvol2
dev 5000 658 4342 13% 0 - 0 /dev/vg00/swap
dev 5116 0 5116 0% 0 - 1 /dev/swapvg/lvol1
reserve - 3773 -3773
memory 3085 332 2753 11%
total 21201 4763 16438 22% - 0 -
=============================================

One more doubt - In the output of swapinfo command, it shows that around 3GB out of 4GB RAM is being reserved for swap. Any comment on this?