jmcdowell@issisolutions.com wrote:
Can anyone please shed some light on this issue?
I know nothing specifically about those clusters, except that they are pretty old by now.
It could be that you're hitting a bug in the old Etherboot which I assume is in the boot flash, when sending it a newer kernel.
One thing you could try is to take out one node into a lab setting and send it various other payloads.
Etherboot should indeed be able to boot anything that is usable as payload with LinuxBIOS/coreboot.
Some suggestions would be FILO, GRUB2, gPXE (yes), and then finally move on to a new kernel.
The tool to use is still mkelfImage, but note that the perl script with the confusingly similar name mkelfimage is NOT useful at least to make a coreboot payload.
The good mkelfImage is available at http://coreboot.org/Mkelfimage
I would send a filo.elf configured to boot a kernel from local storage to Etherboot. Which needs local storage. But I think it's a start for debugging the problem further.
//Peter