Free memory is too little

[Dmitriy_16]

HP cluster which consist of 2 HP servers (RP 5470). System launched to commercial running as real time billing server.
System displays alarm information
(2005-01-19 09:31:33) Warning: ID=7100 INFO=Free Memory is too little
which appeare if Free Memory is less then 20MB.

Average memory usage is 96%-98%. What can be done to decrease memory usage?

Top command displays next
System: scp1 Mon Jan 24 13:49:52 2005
Load averages: 0.08, 0.11, 0.12
144 processes: 136 sleeping, 8 running
Cpu states:
CPU LOAD USER NICE SYS IDLE BLOCK SWAIT INTR SSYS
0 0.16 4.0% 0.0% 0.0% 96.0% 0.0% 0.0% 0.0% 0.0%
1 0.00 1.0% 0.0% 2.0% 97.0% 0.0% 0.0% 0.0% 0.0%
--- ---- ----- ----- ----- ----- ----- ----- ----- -----
avg 0.08 3.0% 0.0% 1.0% 96.0% 0.0% 0.0% 0.0% 0.0%

Memory: 1324508K (1284496K) real, 2001480K (1910984K) virtual, 256548K free Page# 1/4

CPU TTY PID USERNAME PRI NI SIZE RES STATE TIME %WCPU %CPU COMMAND
1 ? 13472 root 152 20 281M 74872K run 0:50 1.17 1.16 java
1 ? 20289 scu 154 20 35440K 23552K sleep 427:50 0.90 0.89 manager
1 ? 5113 scu 154 20 426M 108M sleep 140:02 0.66 0.65 scf
1 ? 5114 scu 154 20 426M 108M sleep 142:03 0.65 0.65 scf
1 ? 5115 scu 154 20 426M 108M sleep 139:51 0.59 0.59 scf
0 ? 35 root 152 20 6464K 6464K run 124:10 0.53 0.53 vxfsd
0 ? 20795 root 158 20 2816K 396K sleep 343:11 0.48 0.48 ksh
1 ? 5112 scu 154 20 518M 224M sleep 148:40 0.40 0.40 scfserver
1 ? 20469 fep 154 20 29128K 17380K sleep 247:42 0.37 0.37 fep
0 ? 16178 scu 154 20 25432K 14468K sleep 211:19 0.32 0.32 oamagent
0 ? 19947 informix 180 20 284M 6184K sleep 138:42 0.27 0.27 oninit
1 ? 19800 root 20 20 11116K 7712K sleep 159:11 0.27 0.27 cmcld
0 ? 1050 root 152 20 15272K 3064K run 1:08 0.23 0.23 dmisp
1 ? 20295 scu 154 20 22516K 9740K sleep 105:09 0.19 0.19 smfagent
0 ? 16196 scu 158 20 2056K 372K sleep 59:33 0.13 0.13 sh
0 ? 20603 ospfep 154 20 11244K 1204K sleep 70:45 0.13 0.13 monfep
1 ? 0 root 127 20 32K 0K sleep 106:06 0.10 0.10 swapper
0 ? 19991 informix 180 20 284M 3628K sleep 35:31 0.09 0.09 oninit
1 ? 19952 informix 180 20 284M 3140K sleep 1:35 0.06 0.06 oninit
1 pts/ta 11273 scu 168 20 2100K 400K sleep 0:00 0.06 0.06 csh
1 ? 19985 informix 180 20 284M 3168K sleep 1:31 0.06 0.06 oninit
1 ? 1 root 168 20 484K 204K sleep 29:50 0.05 0.05 init
1 ? 2321 root 154 10 4472K 500K sleep 42:36 0.05 0.05 psmctd
0 pts/ta 16949 scu 178 20 4612K 572K run 0:00 1.00 0.05 top
0 ? 19980 informix 180 20 284M 3176K sleep 1:34 0.05 0.05 oninit
0 ? 430 root 152 20 1972K 76K run 10:00 0.04 0.04 syncer
1 ? 1059 root 152 20 13028K 1576K run 0:09 0.04 0.04 hpuxci
1 ? 1072 root 152 20 11912K 344K run 0:09 0.04 0.04 swci
1 ? 3 root 128 20 32K 32K sleep 40:58 0.04 0.04 statdaemon
1 ? 1134 root 154 20 2652K 280K sleep 32:13 0.04 0.04 pwgrd
1 ? 19965 informix 180 20 284M 3172K sleep 1:33 0.02 0.02 oninit
1 ? 3006 root 154 20 4196K 336K sleep 1:58 0.02 0.02 dm_stape
0 ? 2 root 128 20 32K 32K sleep 0:24 0.02 0.02 vhand
0 ? 4 root 128 20 32K 32K sleep 1:38 0.02 0.02 unhashdaemon
0 ? 19 root 147 20 32K 32K sleep 3:05 0.02 0.02 lvmkd

[Steve Steel]

Hi

All data about memory usage is at

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


If anything can help you it is there


Steve Steel

[harry d brown jr]

What are your dbc_min/max_pct values?

[root@vpart1 /]# kmtune | grep dbc
dbc_max_pct 10 - 10
dbc_min_pct 5 - 5
[root@vpart1 /]#

live free or die
harry d brown jr

[Dmitriy_16]

# kmtune | grep dbc
dbc_max_pct 4 - 4
dbc_min_pct 5 - 5

[Ravi_8]

Hi,

Increase the values of dbc_max_pct kernel parameter to about 9 or 10 from the current value of 4

[Dmitriy_16]

What means this parameter dbc_max_pct?
and how to change it?
Is it neccessary to restart server after modofication of this parameter?

[bhavin asokan]

hi,
dbc_max_pct is a kernel parameter.

how much is the total physical memory in your system.see the glance output and check howmuch is the buffer cache.


dbc_max_pct
This determines the percentage of main memory to which the dynamic filesystem buffer
cache is allowed to grow (when nbuf and bufpages are zero). The default is 50 percent
of memory, but this is major overkill in most cases. With a huge bufcache, youâ re more
likely to get into a situation where free memory is low and youâ ll need to pageout or
shrink the buffer cache in order to meet memory demands for active processes. You do
not want to get into that situation. If you want to use a dynamic buffer cache, start with
dbc_max_pct at 25. If you have 4 or gigabytes of physical memory, especially on 11.0,
start with it even smaller. We have a subsection below delving more into Buffer Cache issues.
for changing the this parameter you have to rebuild the kernel

see attachment for other kernel parameters

r

[bhavin asokan]

hi,

forgot to attach the file

regds,

[B. Hulst]

Hi,

You are running java and informix.

You can allocate less memory for buffers to informix and reduce the initial min (and max) memory (stack) size for a java virtual machine.
This java memory tuning helps if there are more then one running on your system. I see only one java in this listing but there are 130 processes...

- Bob