topic Re: Kernel Panic in Operating System - Linux

Kernel Panic

Naj — Fri, 08 Jul 2011 01:22:59 GMT

Hi Expert,

Sometime it's difficult to investigate the root cause when receiving error kernel panic. What we did is reboot the server and it back to normal. However, would you share with me is there has any away to trace back/any script execute to translate the error why/what is going on.

Rebooted the server for several times without knowing the root caused is very bad and would be trouble to server for long term,

Thanks in advanced

Naj

Re: Kernel Panic

Steven E. Protter — Mon, 11 Jul 2011 19:52:21 GMT

Shalom,

cd /etc/rc.config.d

crashconf
savecrash

See where there are crash files and if they are configured. Check the locations. There should be a fair number of files that can be tarred up, sent to HP support and analyzed for root casue Analysis.

Please post:

crashconf -v

SEP

Re: Kernel Panic

Alzhy — Wed, 13 Jul 2011 14:28:27 GMT

Greetings and Salam!

So your crashes and "hungs" can be captured for analysis by the vendor, you need to set up your OS properly to capture kernel dump so you can force it to dump kernel or a panic will capture the dump.

First thing is you must have a crashkernel line configured (/boot/grub/menu.lst):

title Red Hat Enterprise Linux Server (2.6.18-194.32.1.el5)
        root (hd0,0)
        kernel /vmlinuz-2.6.18-194.32.1.el5 ro root=/dev/vg00/root fastboot crashkernel=512M@16M elevator=deadline
        initrd /initrd-2.6.18-194.32.1.el5.img

2nd, you must set up /etc/kdump.conf properly to suite your environment. Mine is set up as follows:

#ext3 LABEL=/crash

net netapp01.xyz.com:/crash/linux
core_collector makedumpfile -d 31 -c

Enjoy.!

Re: Kernel Panic

Alzhy — Wed, 13 Jul 2011 14:30:30 GMT

Seriously?
Ain't the above for HPay-UX good sir?

Re: Kernel Panic

Alzhy — Wed, 13 Jul 2011 14:36:46 GMT

I almost forgot, once you have the above set up - whenever your OS panics - it should generate a kernel dump at the location you specified.

If your Servr is behaving badly or seem to be "hung" - You can FORCE your server to generate a crash dump by invoking:

SysRQ+c

It may also help to do several SysRQ+t before doing SysRQ+c so the crash log/dump/etc will capture the various threads running while yor OS is having issues or is "hung"...

Cheers and hope this helps.

Re: Kernel Panic

Alzhy — Wed, 13 Jul 2011 14:43:57 GMT

This is important additional information. I just recalled, for SysRQ+t or +c to work, you must set your kernel parm kernel.sysrq = 1 (enable). You can do this by adding or editing the kernel.sysrq line in /etc/sysctl.conf and comitting the change by doing " sysctl -p".

Your SysRQ+c or +t or whatever SysrQ combo should now work.

Note SysRQ means ALT+SysRq keys. Most virtual consoles of todays X86 servers feature the shortcut already. If you have a physical console -- then its the ALT+SysRQ+c (or t) combination.

Hope this helps.

Re: Kernel Panic

Steven E. Protter — Thu, 14 Jul 2011 14:53:17 GMT

Shalom again,

Was this case in the HP section before? My earlier response made no sense. It was the prefect response to an HP-UX crash.

Is kdump configured? Whats in it? I think you should have a kdump server to accommodate the entire crash.

SEP