How to detect memory leak

[Kurz]

Hi all,

I think we have a memory leak on our SAP DB server. vhand was sometimes as high as 40%, we rebooted, and now it is rising again.

How can I found out which process is consuming the memory?

I attach top and vmstat output. How can I find out more?

best regards,

Michaela

[Suraj K Sankari]

Hi,

>>How can I found out which process is consuming the memory?

oracleC11 is taking more memory see one process is taking 8627M

#ps -aef | grep oracleC11
#ps -aef | grep oracleC11 | wc -l

Your most of the memory is taking by Oracle

I think you need to tune your oracle

Suraj

[Kurz]

Oracle is using shared memory segments, that why the memory consumption appears strange. I.e: It's always the same memory.

best regards,

Michaela

[Jitesh purohit_1]

Hi Michaela ,

The above output does not always shows you have a memory leak on your server..

There are many tools like kmeminfo ( HP ), which can clearly tell if there is a memory leak . So i would suggest you either download & install this tool or log a case with HP Vendor

Analyzing memory performance / Memory leak issues takes lot of understanding of how memory management in HPUX

Thanks
Jitesh

[Michal Kapalka (mikap)]

hi Michaela ,

check the glance or gpm ( gui ) monitoring tool, if you have some bootleneck, a warning message will be generated in other window.

export display where you have X-server and them run gpm.

mikap

[Steven E. Protter]

Shalom,

Use my handy, dandy memory leak detector.

http://www.hpux.ws/?p=8

Works in Solaris, Linux, HP-UX

SEP

[Duncan Edmonstone · Accepted solution]

Michaela,

Large amounts of activity from vhand in 11iv3 do not necessarily mean you have memory pressure. In 11iv3 vhand is also responsible for freeing pages in the file cache. Have a look at the white paper here:

http://h71028.www7.hp.com/ERC/downloads/4AA0-7157ENW.pdf

And Don Morris's comments in this thread:

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1241686

So first things first, make sure you are up-to-date on any VM subsystem patches for 11iv3

From what we see here you aren't getting any VM page-outs, so the chances are the vhand activity is associated with the file cache. As this appears to be an Oracle/SAP system, you should probably ensure that all your Oracle filesystems don't use filecache by setting "minacache=direct,convosync=direct" on your VxFS filesystem mount options (I'm assuming here that you have Online JFS). That might stop the apparent cache thrashing (if thats what's happening and its only a guess based on data provided).

HTH

Duncan