• Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 962
  • Last Modified:

Solaris 9 Swap Leak?

I have a Solaris 9 with Oracle 9.2.0.1 on it. And over a period of time the swap space fills up until completely full.

I though it was Oracle but when I shutdown all the oracle processes it still does not clear the swap usage.


How can I find out what's using the swap space or what is wrong?

Here is the 'top' output after restarting oracle.

last pid:  1123;  load averages:  0.00,  0.01,  0.02                                                                                                 15:23:08
34 processes:  33 sleeping, 1 on cpu
CPU states: 98.7% idle,  0.0% user,  0.7% kernel,  0.6% iowait,  0.0% swap
Memory: 2048M real, 257M free, 3808M swap in use, 1806M swap free

   PID USERNAME THR PRI NICE  SIZE   RES STATE    TIME    CPU COMMAND
  1123 oracle     1  59    0 2632K 1720K cpu/1    0:00  0.34% top
   461 oracle    16  59    0 2940M  700M sleep    4:23  0.05% dbsnmp
  1059 oracle    11  59    0  879M  835M sleep    0:00  0.01% oracle
  1071 oracle     1  59    0  877M  834M sleep    0:09  0.01% oracle
    57 root       9  59    0 3616K 1304K sleep    0:40  0.00% picld
  1069 oracle     1  59    0  878M  840M sleep    0:15  0.00% oracle
  1055 oracle    37  59    0  882M  838M sleep    0:00  0.00% oracle
  1067 oracle     1  59    0  877M  837M sleep    0:00  0.00% oracle
  1061 oracle     1  59    0  877M  837M sleep    0:00  0.00% oracle
  1065 oracle     1  59    0  877M  836M sleep    0:00  0.00% oracle
  1063 oracle     1  59    0  877M  836M sleep    0:00  0.00% oracle
  1057 oracle    11  59    0  883M  836M sleep    0:00  0.00% oracle
  1053 oracle     1  59    0  878M  835M sleep    0:00  0.00% oracle
  1075 oracle     1  59    0  881M  833M sleep    0:00  0.00% oracle
  1073 oracle     1  59    0  881M  833M sleep    0:00  0.00% oracle
  1083 oracle     1  59    0   17M 7048K sleep    0:00  0.00% tnslsnr
   907 root       1  59    0 4608K 2704K sleep    0:00  0.00% sshd
  1107 root      18  59    0 2768K 2088K sleep    0:00  0.00% nscd
   251 root       9  59    0 2912K 1432K sleep    0:00  0.00% syslogd
   254 root       1  59    0 2264K 1184K sleep    0:00  0.00% cron
   909 oracle     1  59    0 1840K 1152K sleep    0:00  0.00% ksh
   381 root       1  59    0 2688K 1016K sleep    0:00  0.00% sshd
   285 root       3  59    0 2880K  912K sleep    0:00  0.00% vold
   472 root       1  59    0 1776K  856K sleep    0:00  0.00% ttymon
   473 root       5  59    0 2688K  824K sleep    0:00  0.00% devfsadm
    52 root      14  59    0 2280K  808K sleep    0:00  0.00% syseventd
   468 root       1  59    0 1768K  808K sleep    0:00  0.00% sac
   296 root       1  59    0 2664K  792K sleep    0:00  0.00% mdmonitord
   243 root       1  59    0 1952K  776K sleep    0:00  0.00% inetd
   469 root       1  59    0 1776K  760K sleep    0:00  0.00% ttymon


I could only find this on sun's site, http://docs.sun.com/db/doc/806-5195/6je7ls071?a=view#runtimebugs-1147

But would rather find out what's causing it it before changing any config.
0
grahamcruickshanks
Asked:
grahamcruickshanks
  • 2
1 Solution
 
beskyCommented:
If you have that CDE-packages loaded, run
sdtprocess and display on a ws running X.
Here you can see each process swapallocation ( and more)

If not, run pmap -a on the processes you suspect.
It will show you swapallocation by proc.


There is a Memtool application on the net which is quite nice and a prtswap utility.

Do a search on the net and you'll find it available in your area.

But I dont think any of the Solris-processes are leaking,
this would for sure have triggered a bugreport

HTH
0
 
grahamcruickshanksAuthor Commented:
I shut down oracle still all the memory was inuse. Then I noticed after I posted this question.

461 oracle    16  59    0 2940M  700M sleep    4:23  0.05% dbsnmp

was not shutdown.

killed that process and all my memory was freed up!!!.

I don't need dbsnmp running as I'm not using a Oracle management server for backups.

So I've disabled it and everything is seems much better.

0
 
grahamcruickshanksAuthor Commented:
Futher Info:

Using MemTool 3.9.4 from

ftp://playground.sun.com/pub/memtool

It has a command 'mem' which displays a VFS Memory Dump

giving a list of file memory for processes. (swap usage)

Also page in/outs

0

Featured Post

[Webinar On Demand] Database Backup and Recovery

Does your company store data on premises, off site, in the cloud, or a combination of these? If you answered “yes”, you need a data backup recovery plan that fits each and every platform. Watch now as as Percona teaches us how to build agile data backup recovery plan.

  • 2
Tackle projects and never again get stuck behind a technical roadblock.
Join Now