Red Hat

Dear All ,

One of our Oracle Database Server , which got hanged a couple of days before , so that we could not login to the Server , but it was pinging.
So we tried various options and at last we did a reboot of this server.
To find out the root cause of the Server Hang , when i checked the logs in /var/log /messages , i dont find any entry in the period it got hanged and logs are updated after the restart.

To find out the RAM , CPU Usage also , i checked sar outputs , there also i dont find any entry for the day it got hanged.

Now i want to know how to find the root cause for this.

Can any one pl help me.

little silly things can lead to such issues you know, like an employee printing on a friday 17:00 but decides its weekend and so leave before it had time to print and swithes off the printer and leaves... the machine load will inevitably start to raise with all the attempts to print...
You dont give us enough information for us to give you our thoughts only to say yes it can happen like my example (happened to me years ago...) but also "dead processes" big cache flush can take quite some time etc... And a faulty network interface: on a AIX I had once a card that would ping but not much more haha

Code:

dmesg | grep -i killed

What's the status of memory overcommit?

Code:

echo /proc/sys/vm/overcommit_memory

For a database server, memory overcommit should be disabled:

Code:

sysctl vm.overcommit_memory=2
echo "vm.overcommit_memory=2" >> /etc/sysctl.conf

You do not want to overcommit memory on a server where long-lived processes WILL use the memory they allocate.

Otherwise the OOM killer will kill important processes, possibly hanging the machine. And when a machine is hung badly, it may not even be able to log anything or record SAR data.

If you find your database or other processes no longer start up after you disable memory overcommit, you were using too much memory in the first place. Buy more RAM and/or add swap.