How to take memory dump in Linux

rhel1984 · 09-26-2012, 12:02 AM

Hi All,

I need to take memory dump from the server(rhel 5) for troubleshooting purpose.
My rhel server is very much up and running. I tried to search the existing threads but was not of much help.
Can you please help me in this.

Thanks

MensaWater · 09-26-2012, 09:09 AM

Are you asking how to force a panic to make a kdump crash dump? If so read the warning and do the command below.

***WARNING***
The command below crashes your running system (forces a reboot).
This could have consequences for your filesystems and especially for any databases you might be running.
***WARNING***

Code:

echo c > /proc/sysrq-trigger

The dump will be put in /var/crash after the reboot.

rhel1984 · 09-26-2012, 10:56 PM

Thanks for the reply.
The server is in production environment so I cant go for crash dump.
We are facing high memory usage issues and in-order to troubleshoot the issue we need to take
memory dump for a specific application process.
Is there any way to do this without rebooting?

Thanks..

MensaWater · 09-27-2012, 11:41 AM

You can run strace on the actual process.
Type "man strace" for more details.

You can also look at memory maps and other detail for the process under the /proc directory structure.
Type "man proc" for more details.

Ginola · 10-01-2012, 08:08 AM

Out of interest, what is the Application?

cat /proc/[pid]/maps, then gdb works for us.