system error

[juno2]

Sometimes when we run database process ,then it will pop up the error "SYSTEM ERROR: Memory violation. " , I check that the error may due to the following reason :

1) Not enough memory or corrupted memory on the machine
2) Not enough configured swap space on the machine
3) Hardware failure

How can I fix the error? Thx.

[monasingh_1]

hardware failure (any memory bit error) will automatically be logged in /var/adm/syslog/syslog.log

for swaputilization, you can run swapinfo , if you are at 100% utilization, you need to increase swap.

If the error is coming when you run svrmgrl itself to start the database that will mean the you need to increase shmmax value..

[Sridhar Bhaskarla]

Hi,

What is the database process?. Is it 32bit or 64bit?. If it is 32bt and if you are not able to start the database means it could be due to insufficient/fragmented shared memory. If you bring down all the application and start them in order, then you may get your database running.

To find out if it is due to swap do a "swapinfo -t" and observe the total column. If it is around 100%, then your system is running out of virtual memory and you will need to add more swap space. Find out a volume group that is idle and has space in it.

#lvcreate -n lvswap -L size_in_mb -r N vgxx

vi /etc/fstab and keep the entry

/dev/vgxx/lvswap ... swap pri=0 0 0 and then run

#swapon -a

I have not seen hardware failures giving out these messages. As long as the failure is an LPMC, system will take care of it. If it a HPMC, system will crash.

-Sri

[juno2]

Except increase the swap memory ( cos the system can't be shutdown ) , to temporary solve the problem , can I limit the memory /CPU usage of the each process ? thx.

[Sridhar Bhaskarla]

Hi,

It is not possible to limit the cpu for each process. However, memory limitation is it depends. The parameters maxdsiz, maxssiz, maxtsiz and shmmax control limits on various memory regions. You will need to use PRM/WLM (process resource manager or work load manager) to literally control the resources.

To increase the swap, you do not need to shutdown the box. Just follow the process I outlined above. You can do it online.

To decrease swap however, you will need to reboot.

-Sri

[juno2]

What is the pitfall to increase swap memory ? thx.

[monasingh_1]

just a note that if your maxswapchunks kernel parameter is just enough to support your current swap, then you will have to reboot to add any extra swap..

one thing you can also try is running ipcs
and see if any user is occupying any shared mem segment which whould not be present, changces are less if all programs are running for production purpose.

If your box is 11i then few memory parameters are dynamic which you can change on the fly and try again to see if the problem is gone...

hope this helps..