Memory bottleneck situation

[nightwich]

Hi Gurus.

I have a machine:

HP-UX Server_name B.11.31 U 9000/800 3010439677 unlimited-user license

Physical memory: 16Gb

swapinfo -tam:
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 24544 8 24536 0% 0 - 1 /dev/vg00/lvol2
reserve - 10529 -10529
memory 15551 12258 3293 79%
total 40095 22795 17300 57% - 0 -


vmstat 5 5:
procs memory page faults cpu
r b w avm free re at pi po fr de sr in sy cs us sy id
2 0 0 720389 18304 0 0 6 3 0 0 0 3642 19907 574 8 2 90
2 0 0 720389 17871 0 0 0 0 0 0 0 3692 26266 676 17 3 80
6 3 0 770966 17599 0 0 21 0 0 0 0 5150 55350 1345 20 10 70
6 3 0 770966 17410 0 0 6 0 0 0 0 5289 40879 1243 15 3 82
2 0 0 757191 17354 0 0 0 0 0 0 0 3949 25049 830 14 2 83


top:

System Page Size: 4Kbytes
Memory: 9341248K (2367872K) real, 11261136K (2874496K) virtual, 70424K free Page# 1/35

GPM: attached


Is my memory under stress, I can see I am out of physical memory only near 100mb left, but as you can see I do not use eat any swap memory or having page outs.

Can any one put their inputs about this ?

I will be generous :) !!..

Regards.

[Aneesh Mohan]

Yes, your system is under memory bottle neck.

Please post o/p of
#UNIX95=1 ps -e -f -o vsz,sz,ruser,uid,pid,args |sort -rnk1 |head -n 30

Aneesh

[nightwich]

Hi Mohan,

The output of UNIX95=1 ps -e -f -o vsz,sz,ruser,uid,pid,args |sort -rnk1 |head -n 30


root@name_server:/# UNIX95=1 ps -e -f -o vsz,sz,ruser,uid,pid,args |sort -rnk1 |head -n 30
180800 22804 oracle11 109 23355 oracleWSPROD (LOCAL=NO)
180800 22022 oracle11 109 27330 oracleWSPROD (LOCAL=NO)
172608 21628 oracle11 109 17904 oracleWSPROD (LOCAL=NO)
172608 21616 oracle11 109 2682 oracleWSPROD (LOCAL=NO)
172608 21503 oracle11 109 23377 oracleWSPROD (LOCAL=NO)
172608 21315 oracle11 109 29630 oracleWSPROD (LOCAL=NO)
172608 21309 oracle11 109 18726 oracleWSPROD (LOCAL=NO)
172608 21290 oracle11 109 12980 oracleWSPROD (LOCAL=NO)
172608 21262 oracle11 109 29311 oracleWSPROD (LOCAL=NO)
172608 21246 oracle11 109 11194 oracleWSPROD (LOCAL=NO)
172608 21236 oracle11 109 24571 oracleWSPROD (LOCAL=NO)
172608 21236 oracle11 109 13438 oracleWSPROD (LOCAL=NO)
172608 21198 oracle11 109 18618 oracleWSPROD (LOCAL=NO)
172608 21180 oracle11 109 12427 oracleWSPROD (LOCAL=NO)
172608 21152 oracle11 109 24389 oracleWSPROD (LOCAL=NO)
172608 21081 oracle11 109 12748 oracleWSPROD (LOCAL=NO)
172608 21053 oracle11 109 12921 oracleWSPROD (LOCAL=NO)
172608 21027 oracle11 109 29602 oracleWSPROD (LOCAL=NO)
172608 20957 oracle11 109 7386 oracleWSPROD (LOCAL=NO)
172608 20956 oracle11 109 4587 oracleWSPROD (LOCAL=NO)
172608 20933 oracle11 109 16472 oracleWSPROD (LOCAL=NO)
172608 20929 oracle11 109 4523 oracleWSPROD (LOCAL=NO)
172608 20924 oracle11 109 12159 oracleWSPROD (LOCAL=NO)
172608 20918 oracle11 109 717 oracleWSPROD (LOCAL=NO)
172608 20915 oracle11 109 12935 oracleWSPROD (LOCAL=NO)
172608 20898 oracle11 109 29968 oracleWSPROD (LOCAL=NO)
172608 20894 oracle11 109 19313 oracleWSPROD (LOCAL=NO)
172608 20888 oracle11 109 25960 oracleWSPROD (LOCAL=NO)
172608 20878 oracle11 109 29048 oracleWSPROD (LOCAL=NO)
172608 20867 oracle11 109 21244 oracleWSPROD (LOCAL=NO)


Thanks..

[Aneesh Mohan]

What I can see from the above o/p is
You have an Oracle instance WSPROD with more PGA allocation.

Escalate this to ORACLE DBA , they may change dedicated server process to shared or they will tune the PGA (Process Global Area)

Aneesh

[nightwich]

Hi,

ok. I want to explore the O.S. side.


In the glance I can see that the system memory is taking 5GB is this normal ?

Can I tune this ?.

What ware the options here. More physical memorie more swap ?.


Thanks in advance !! ..

[DeafFrog]

HI ,

can you please post the o/p of

#kctune filecache_max
#kctune filecache_min
#kctune vxfs_ifree_timelag
#kctune vx_ninode


Regards ,

[nightwich]

Hi


The output of kctune:

root@name_server:/# kctune filecache_min
Tunable Value Expression Changes
filecache_min 815297331 5% Imm (auto disabled)
root@name_server:/# kctune vxfs_ifree_timelag
Tunable Value Expression Changes
vxfs_ifree_timelag 0 Default Immed
root@name_server:/# kctune vx_ninode
Tunable Value Expression Changes
vx_ninode 0 Default Immed


Regards

[Aneesh Mohan]

Please post ,

UNIX95=1 ps -e -f -o vsz,sz,ruser,uid,pid,args |grep root|sort -rnk1 |head -n 30


Aneesh

[DeafFrog]

Hi ,

root@name_server:/# kctune filecache_max ?

Regards ,

[Michael Steele_2 · Accepted solution]

Hi

a) Your swap is at 57% and fine.

b) vsz is used to find memory leaks, not memory bottlenecks and since all your processes' are very similar and associated to oracle, I would contact the oracle DBA about the size of the oracle SGA. Note: To find a memory leak you'll need to gather 2-3 days of vsz info and observe any vsz growth.

c) 5 GB system memory is fine.