Memory problems

[Robert Dukes]

Hi,

We are seeing memory utilisation problems on an rp7410 with 8Gb of memory running Oracle. I have posted a couple of threads related to this issue in this area and have had a great response. Could anyone tell me if it is possible to identify whether a process is locking itself into real memory? Is there a command to do this? We believe that this might be the cause of our memory issues.

Thanks in advance for any help.

Rob.

[Roberto Arias]

Good afternoon:


please try this:

ipsc -m (for look for process in shared memory)
iprm -m

and
ipcs -s (for look for semaphores)
iprm -s


best regards

[Eric Antunes]

Hi Robert,

My experience tells me that most memory issues are related with bad SQL witch is sometimes due to bad analisys.

One way to see bad sql is querying the huge disk_reads (more than 1000):

select sql_text, executions, disk_reads, loads, rows_processed, optimizer_mode, module
from v$sqlarea
where disk_reads >= 1000

This query also depends of how long SQL will remain in you sga...

Consider not just the disk_reads value here, but the disk_reads percentage for each execution.

Best Regards,

Eric Antunes

[Victor BERRIDGE]

Hi Robert,

What are the applications you are running? this may help us to do a bit of brainstorming...

The issues I had with memory always had something to do with java somewhere(oracle, pvcs...).
The way I try to solve?
easy way: All that is not delivered with the box is launched in rc4.d so I do after asking the concerned folk (but they are generally stuck and awaiting for a solution...) if they can stop the applications or I do an init 3.
Then I look whats left: ipcs -a.
And start cleaning: ipcrm....
but I write down the owners and then try to see what could have happened.

So if you suspect some apps not releasing memory after usage I can only see that way of doing


All the best
Victor

[Sridhar Bhaskarla]

Hi Rob,

Run

UNIX95= ps -e -o 'vsz pid args' |sort -n

periodically. It reports the processes sorted by their relative memory usage.

If I remember correctly, I suggested you to add a 2GB of swap space. I believe it's not the memory issue but a mere 'swap space' issue.

-Sri

[Kent Ostby]

Robert --

HP in general recommends that you have at least 2 x MEMORY worth of Device Swap.

I have seen hang dumps on Oracle engines that required the user to approach 3 x MEMORY worth of Device Swap.

I would definitely tune this as you approach your problem as well.

Best regards,

Oz

[Geoff Wild]

I read your other thread on swap - did you change your DBC_MAX_PCT kernel parameter yet?

ipcs -ma to see shared memory regions.

Rgds...Geoff