Clueless about Kdump
 

Hi,

I have 2 Linux clusters....so 4 machines.
The first 2 machines (1&2) are functioning fine, 3 & 4 are having some problems in that one of them has randomly crashed 2 or 3 times over the course of the last 2 weeks.

I was told to check for "kdumps". I read up on the KDump utility and fwiw machines 1 & 2 have the /usr/lib/debug directory and stepping down through this directory brings me to a file "vmlinux". Machines 3 & 4 have no such directory. Does that mean that Kdump is not installed/setup on 3 and 4?
How can I tell? And how can I activate it?

Also, does anyone have a clue how to analyse kdump files? (but that's probably a separate issue. Right now I just need to make sure that machines 3 & 4 have some auditing done for the next time one of them falls over.

Thank you.

Lets start with the basics; what distro+version are on each of the machines (post results for all 4)?

Hi Chrism,

OK.....back to basics.
I recently tried to configure kdump on a RH box ( 2.6.18-238.el5 ).
#service kdump status shows kdump to be operational.

I sent a crash signal to the system (echo c > /proc/sysrq-trigger)

but when the machine comes back there's no vmcore in the /var/crash directory.

I'm certain that I've missed a very important part of the setup like kernel parameters but the documentation is unclear.

Thanks for any help.