High page out - What is this, how can I solve this.

[haneps]

Hi Guys,

Good day to all.

I have a high "page out" on my server. The vmstat output below tells me that pageout count is 35. I heard from a friend that it should not be higher than 20.

Can someone enlighten me what is page out and how may I decrease the amount to below 20.



Thanks.

-+---------------------------------------------------------------------------------------------------+-
# vmstat
procs memory page faults cpu
r b w avm free re at pi po fr de sr in sy cs us sy id
0 0 0 259114 384072 98 44 41 35 59 0 0 4424 7357 1215 16 7 77
#
-+---------------------------------------------------------------------------------------------------+-

# uname -am
HP-UX xxx B.11.00 U 9000/800 137444646 unlimited-user license
#

[Steven E. Protter]

Shalom,

Supply and demand.

Increase the supply of memory.

Or.

Decrease the demand for it by running less programs or making your Oracle SGA's demand less memory.

Usually I just buy more memory.

SEP

[Pete Randall]

Actually, po is rarely a problem, it's the pi that gets you. Since your pi is high as well, you've got problems. As Steve suggests, you need to either aquire more or free some existing memory. What are you kernel settings for dbc_max_pct and dbc_min_pct set to? The default values generally consume far too much memory.


Pete

[Pete Randall]

Sorry - that was backwards: pi is rarely a problem, po IS.

[haneps]

My memory uses are not that high.

# swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 512 0 512 0% 0 - 1 /dev/vg00/lvol2
dev 8192 4 8188 0% 0 - 0 /dev/vg00/lvol9
reserve - 3024 -3024
memory 3417 534 2883 16%
total 12121 3562 8559 29% - 0 -
#

Attached is my memory allocation on the processes.

Thanks.

[Bill Hassell]

From your attachment:

> # UNIX95=1 ps -e -o vsz,ruser,pid,args | sort -rn
> 917368 root 1092 /usr/sbin/mib2agt

Whoops, runaway mib2agt...well known problem fixed by patches for 11.00 or better yet, kill it and then turn off all Snmp processes (unless you are required to have them running by your network operations center). Snmp is not needed otherwise. Be sure to turn off all Snmp startup scripts:

/etc/rc.config.d/SnmpAtm*
/etc/rc.config.d/SnmpHpunix
/etc/rc.config.d/SnmpIpv6
/etc/rc.config.d/SnmpMaster
/etc/rc.config.d/SnmpMib2
/etc/rc.config.d/SnmpTrpDst

Inside each script (you may not have all of these at 11.00), there is a flag to start the service. Set them all to 0 and the next reboot will not have any SNMP services running.

A constant page out in double digits could be normal if you know that your application(s) are making extensive use of memory mapped files. It appears you have plenty of memory so the po value probably does not refer to actual process pageout.