Swap memory 100%

[Prashantj]

Hi,

We have 11.11 hp-ux production server. The server taking to much memory usage 100%.
we already analyse the system and we did't found any processs which is taking high usage.

#swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4096 371 3725 9% 0 - 1 /dev/vg00/lvol2
reserve - 794 -794
memory 1531 1481 50 97%
total 5627 2646 2981 47% - 0 -

Pls suggest if we require to change belw kernal parameter.

#kmtune | grep dbc_max_pct
dbc_max_pct 50 - 50

Thanks
Prashant

[Ganesan R]

Hi Prashant,

Are you talking about swap memory or physical memory. Note that the "memory" column on the swapinfo output not the actual memory usage. It represents the pseduo memory. For more details about pseduo read this.

ftp://hprc.external.hp.com/memory.htm#the_memory_line_in_the_output_of_swapinfo

To know your physical memory usage use glance or kmeminfo tool. And dbc_max_pct is set to default 50% which is not needed normally. You can reduce it to 10-15% depends on your application requirement.

Hope this helps.

[James R. Ferguson]

Hi:

Aside from adding more memory, you probably should reduce your buffer cache. A value somewhere between 5-10% is probably satisfactory. Using 'sar -b' to monitor buffer cache hit ratios may provide the empirical evidence you need.

Regards!

...JRF...

[Prashantj]

Hi

Thanks for your reply.

i am talking about physical memory.

Unable to connect. any other way.
ftp://hprc.external.hp.com/memory.htm#the_memory_line_in_the_output_of_swapinfo

Glance Output
B3692A GlancePlus C.03.71.00 09:58:29 hbashp01 9000/800 Current Avg High
-----------------------------------------------------------------------------------------------------------------------------
CPU Util S SUU | 7% 6% 12%
Disk Util F FV V | 46% 27% 46%
Mem Util S SU UB B |100% 100% 100%
Swap Util U UR R | 47% 47% 47%
-----------------------------------------------------------------------------------------------------------------------------
PROCESS LIST Users= 90
User CPU Util Cum Disk Thd
Process Name PID PPID Pri Name ( 200% max) CPU IO Rate RSS Cnt
--------------------------------------------------------------------------------
swapper 0 0 127 root 1.7/ 1.7 0.2 10.9/ 7.8 32kb 1
uniapisvr 19152 14838 152 root 1.3/ 1.3 0.1 7.3/ 7.3 32kb 1
uniapisvr 19155 19153 152 swebhm 1.1/ 1.1 0.1 3.6/ 3.6 32kb 1
vhand 2 0 128 root 0.8/ 0.7 0.1 259/ 141 32kb 1
uniapisvr 19153 14838 178 swebhm 0.8/ 0.8 0.0 1.9/ 1.9 992kb 1
sh 19158 19152 179 root 0.2/ 0.2 0.0 2.1/ 2.1 308kb 1
udt 18814 1 168 cmi 0.0/ 0.0 0.0 4.6/ 3.3 1.7mb 1
vxfsd 35 0 138 root 0.0/ 0.0 0.0 12.6/13.0 1.8mb 30
udt 28197 28119 154 BWIMPEE 0.0/ 0.0 0.0 1.5/ 0.6 2.5mb 1

Memory Output in glance


Total VM : 1.26gb Sys Mem : 764.1mb User Mem: 471.1mb Phys Mem: 2.00gb
Active VM: 617.6mb Buf Cache: 808.3mb Free Mem: 4.6mb

Thanks
Prashant

[Ganesan R · Accepted solution]

Hi Prashant,

If that link doesn't work, try this.

http://docs.google.com/View?docid=dc7s6mzr_56cgsg9g

Also read this thread..
http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1288219


Yes. Your system is having memory crunch. First option is, reduce the buffer cache. It will release your 40% of your physical memory. Then use tools like kmeminfo or glance to find which process is consuming more memory and act accordingly.

[Jeff Gyurko]

You're using 9% of swap. Not terrible, but there are those that say any swap being used indicates more memory is needed.

Judging by the process names, you're running an Infolease product, am I right? If this is the case, then the dbc_max_pct should be reduced to 10 at the most since the database is comprised of flat files and this setting only takes away available memory.